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 !defined(SOLARIS) && defined(_KERNEL) && !defined(__osf__) && \
107 !defined(__hpux)
108 # include <netinet6/in6_var.h>
109 # endif
110 #endif
111 #include <netinet/tcpip.h>
112 #include "netinet/ip_fil.h"
113 #include "netinet/ip_nat.h"
114 #include "netinet/ip_frag.h"
115 #include "netinet/ip_state.h"
116 #include "netinet/ip_proxy.h"
117 #include "netinet/ip_auth.h"
118 #include "netinet/ipf_stack.h"
119 #ifdef IPFILTER_SCAN
120 # include "netinet/ip_scan.h"
121 #endif
122 #ifdef IPFILTER_SYNC
123 # include "netinet/ip_sync.h"
124 #endif
125 #include "netinet/ip_pool.h"
126 #include "netinet/ip_htable.h"
127 #ifdef IPFILTER_COMPILED
128 # include "netinet/ip_rules.h"
129 #endif
130 #if defined(IPFILTER_BPF) && defined(_KERNEL)
131 # include <net/bpf.h>
132 #endif
133 #if defined(__FreeBSD_version) && (__FreeBSD_version >= 300000)
134 # include <sys/malloc.h>
135 # if defined(_KERNEL) && !defined(IPFILTER_LKM)
136 # include "opt_ipfilter.h"
137 # endif
138 #endif
139 #include "netinet/ipl.h"
140 #if defined(_KERNEL)
141 #include <sys/sunddi.h>
142 #endif
143 /* END OF INCLUDES */
144
145 #if !defined(lint)
146 static const char sccsid[] = "@(#)fil.c 1.36 6/5/96 (C) 1993-2000 Darren Reed";
147 static const char rcsid[] = "@(#)$Id: fil.c,v 2.243.2.64 2005/08/13 05:19:59 darrenr Exp $";
148 #endif
149
150 #ifndef _KERNEL
151 # include "ipf.h"
152 # include "ipt.h"
153 # include "bpf-ipf.h"
154 extern int opts;
155
156 # define FR_VERBOSE(verb_pr) verbose verb_pr
157 # define FR_DEBUG(verb_pr) debug verb_pr
158 #else /* #ifndef _KERNEL */
159 # define FR_VERBOSE(verb_pr)
160 # define FR_DEBUG(verb_pr)
161 #endif /* _KERNEL */
162
163
164 char ipfilter_version[] = IPL_VERSION;
165 int fr_features = 0
166 #ifdef IPFILTER_LKM
167 | IPF_FEAT_LKM
168 #endif
169 #ifdef IPFILTER_LOG
170 | IPF_FEAT_LOG
171 #endif
172 #ifdef IPFILTER_LOOKUP
173 | IPF_FEAT_LOOKUP
174 #endif
175 #ifdef IPFILTER_BPF
176 | IPF_FEAT_BPF
177 #endif
178 #ifdef IPFILTER_COMPILED
179 | IPF_FEAT_COMPILED
180 #endif
181 #ifdef IPFILTER_CKSUM
182 | IPF_FEAT_CKSUM
183 #endif
184 #ifdef IPFILTER_SYNC
185 | IPF_FEAT_SYNC
186 #endif
187 #ifdef IPFILTER_SCAN
188 | IPF_FEAT_SCAN
189 #endif
190 #ifdef USE_INET6
191 | IPF_FEAT_IPV6
192 #endif
193 ;
194
195 #define IPF_BUMP(x) (x)++
196
197 static INLINE int fr_ipfcheck __P((fr_info_t *, frentry_t *, int));
198 static INLINE int fr_ipfcheck __P((fr_info_t *, frentry_t *, int));
199 static int fr_portcheck __P((frpcmp_t *, u_short *));
200 static int frflushlist __P((int, minor_t, int *, frentry_t **,
201 ipf_stack_t *));
202 static ipfunc_t fr_findfunc __P((ipfunc_t));
203 static frentry_t *fr_firewall __P((fr_info_t *, u_32_t *));
204 static int fr_funcinit __P((frentry_t *fr, ipf_stack_t *));
205 static INLINE void frpr_ah __P((fr_info_t *));
206 static INLINE void frpr_esp __P((fr_info_t *));
207 static INLINE void frpr_gre __P((fr_info_t *));
208 static INLINE void frpr_udp __P((fr_info_t *));
209 static INLINE void frpr_tcp __P((fr_info_t *));
210 static INLINE void frpr_icmp __P((fr_info_t *));
211 static INLINE void frpr_ipv4hdr __P((fr_info_t *));
212 static INLINE int frpr_pullup __P((fr_info_t *, int));
213 static INLINE void frpr_short __P((fr_info_t *, int));
214 static INLINE void frpr_tcpcommon __P((fr_info_t *));
215 static INLINE void frpr_udpcommon __P((fr_info_t *));
216 static INLINE int fr_updateipid __P((fr_info_t *));
217 #ifdef IPFILTER_LOOKUP
218 static int fr_grpmapinit __P((frentry_t *fr, ipf_stack_t *));
219 static INLINE void *fr_resolvelookup __P((u_int, u_int, lookupfunc_t *,
220 ipf_stack_t *));
221 #endif
222 static void frsynclist __P((int, int, void *, char *, frentry_t *,
223 ipf_stack_t *));
224 static void *fr_ifsync __P((int, int, char *, char *,
225 void *, void *, ipf_stack_t *));
226 static ipftuneable_t *fr_findtunebyname __P((const char *, ipf_stack_t *));
227 static ipftuneable_t *fr_findtunebycookie __P((void *, void **, ipf_stack_t *));
228
229 /*
230 * bit values for identifying presence of individual IP options
231 * All of these tables should be ordered by increasing key value on the left
232 * hand side to allow for binary searching of the array and include a trailer
233 * with a 0 for the bitmask for linear searches to easily find the end with.
234 */
235 const struct optlist ipopts[20] = {
236 { IPOPT_NOP, 0x000001 },
237 { IPOPT_RR, 0x000002 },
238 { IPOPT_ZSU, 0x000004 },
239 { IPOPT_MTUP, 0x000008 },
240 { IPOPT_MTUR, 0x000010 },
241 { IPOPT_ENCODE, 0x000020 },
242 { IPOPT_TS, 0x000040 },
243 { IPOPT_TR, 0x000080 },
244 { IPOPT_SECURITY, 0x000100 },
245 { IPOPT_LSRR, 0x000200 },
246 { IPOPT_E_SEC, 0x000400 },
247 { IPOPT_CIPSO, 0x000800 },
248 { IPOPT_SATID, 0x001000 },
249 { IPOPT_SSRR, 0x002000 },
250 { IPOPT_ADDEXT, 0x004000 },
251 { IPOPT_VISA, 0x008000 },
252 { IPOPT_IMITD, 0x010000 },
253 { IPOPT_EIP, 0x020000 },
254 { IPOPT_FINN, 0x040000 },
255 { 0, 0x000000 }
256 };
257
258 #ifdef USE_INET6
259 struct optlist ip6exthdr[] = {
260 { IPPROTO_HOPOPTS, 0x000001 },
261 { IPPROTO_IPV6, 0x000002 },
262 { IPPROTO_ROUTING, 0x000004 },
263 { IPPROTO_FRAGMENT, 0x000008 },
264 { IPPROTO_ESP, 0x000010 },
265 { IPPROTO_AH, 0x000020 },
266 { IPPROTO_NONE, 0x000040 },
267 { IPPROTO_DSTOPTS, 0x000080 },
268 { 0, 0 }
269 };
270 #endif
271
272 struct optlist tcpopts[] = {
273 { TCPOPT_NOP, 0x000001 },
274 { TCPOPT_MAXSEG, 0x000002 },
275 { TCPOPT_WINDOW, 0x000004 },
276 { TCPOPT_SACK_PERMITTED, 0x000008 },
277 { TCPOPT_SACK, 0x000010 },
278 { TCPOPT_TIMESTAMP, 0x000020 },
279 { 0, 0x000000 }
280 };
281
282 /*
283 * bit values for identifying presence of individual IP security options
284 */
285 const struct optlist secopt[8] = {
286 { IPSO_CLASS_RES4, 0x01 },
287 { IPSO_CLASS_TOPS, 0x02 },
288 { IPSO_CLASS_SECR, 0x04 },
289 { IPSO_CLASS_RES3, 0x08 },
290 { IPSO_CLASS_CONF, 0x10 },
291 { IPSO_CLASS_UNCL, 0x20 },
292 { IPSO_CLASS_RES2, 0x40 },
293 { IPSO_CLASS_RES1, 0x80 }
294 };
295
296
297 /*
298 * Table of functions available for use with call rules.
299 */
300 static ipfunc_resolve_t fr_availfuncs[] = {
301 #ifdef IPFILTER_LOOKUP
302 { "fr_srcgrpmap", fr_srcgrpmap, fr_grpmapinit },
303 { "fr_dstgrpmap", fr_dstgrpmap, fr_grpmapinit },
304 #endif
305 { "", NULL }
306 };
307
308
309 /*
310 * Below we declare a list of constants used only by the ipf_extraflush()
311 * routine. We are placing it here, instead of in ipf_extraflush() itself,
312 * because we want to make it visible to tools such as mdb, nm etc., so the
313 * values can easily be altered during debugging.
314 */
315 static const int idletime_tab[] = {
316 IPF_TTLVAL(30), /* 30 seconds */
317 IPF_TTLVAL(1800), /* 30 minutes */
318 IPF_TTLVAL(43200), /* 12 hours */
319 IPF_TTLVAL(345600), /* 4 days */
320 };
321
322
323 /*
324 * The next section of code is a a collection of small routines that set
325 * fields in the fr_info_t structure passed based on properties of the
326 * current packet. There are different routines for the same protocol
327 * for each of IPv4 and IPv6. Adding a new protocol, for which there
328 * will "special" inspection for setup, is now more easily done by adding
329 * a new routine and expanding the frpr_ipinit*() function rather than by
330 * adding more code to a growing switch statement.
331 */
332 #ifdef USE_INET6
333 static INLINE int frpr_ah6 __P((fr_info_t *));
334 static INLINE void frpr_esp6 __P((fr_info_t *));
335 static INLINE void frpr_gre6 __P((fr_info_t *));
336 static INLINE void frpr_udp6 __P((fr_info_t *));
337 static INLINE void frpr_tcp6 __P((fr_info_t *));
338 static INLINE void frpr_icmp6 __P((fr_info_t *));
339 static INLINE void frpr_ipv6hdr __P((fr_info_t *));
340 static INLINE void frpr_short6 __P((fr_info_t *, int));
341 static INLINE int frpr_hopopts6 __P((fr_info_t *));
342 static INLINE int frpr_routing6 __P((fr_info_t *));
343 static INLINE int frpr_dstopts6 __P((fr_info_t *));
344 static INLINE int frpr_fragment6 __P((fr_info_t *));
345 static INLINE int frpr_ipv6exthdr __P((fr_info_t *, int, int));
346
347
348 /* ------------------------------------------------------------------------ */
349 /* Function: frpr_short6 */
350 /* Returns: void */
351 /* Parameters: fin(I) - pointer to packet information */
352 /* */
353 /* IPv6 Only */
354 /* This is function enforces the 'is a packet too short to be legit' rule */
355 /* for IPv6 and marks the packet with FI_SHORT if so. See function comment */
356 /* for frpr_short() for more details. */
357 /* ------------------------------------------------------------------------ */
frpr_short6(fin,xmin)358 static INLINE void frpr_short6(fin, xmin)
359 fr_info_t *fin;
360 int xmin;
361 {
362
363 if (fin->fin_dlen < xmin)
364 fin->fin_flx |= FI_SHORT;
365 }
366
367
368 /* ------------------------------------------------------------------------ */
369 /* Function: frpr_ipv6hdr */
370 /* Returns: Nil */
371 /* Parameters: fin(I) - pointer to packet information */
372 /* */
373 /* IPv6 Only */
374 /* Copy values from the IPv6 header into the fr_info_t struct and call the */
375 /* per-protocol analyzer if it exists. */
376 /* ------------------------------------------------------------------------ */
frpr_ipv6hdr(fin)377 static INLINE void frpr_ipv6hdr(fin)
378 fr_info_t *fin;
379 {
380 ip6_t *ip6 = (ip6_t *)fin->fin_ip;
381 int p, go = 1, i, hdrcount;
382 fr_ip_t *fi = &fin->fin_fi;
383
384 fin->fin_off = 0;
385
386 fi->fi_tos = 0;
387 fi->fi_optmsk = 0;
388 fi->fi_secmsk = 0;
389 fi->fi_auth = 0;
390
391 p = ip6->ip6_nxt;
392 fi->fi_ttl = ip6->ip6_hlim;
393 fi->fi_src.in6 = ip6->ip6_src;
394 fi->fi_dst.in6 = ip6->ip6_dst;
395 fin->fin_id = 0;
396
397 hdrcount = 0;
398 while (go && !(fin->fin_flx & (FI_BAD|FI_SHORT))) {
399 switch (p)
400 {
401 case IPPROTO_UDP :
402 frpr_udp6(fin);
403 go = 0;
404 break;
405
406 case IPPROTO_TCP :
407 frpr_tcp6(fin);
408 go = 0;
409 break;
410
411 case IPPROTO_ICMPV6 :
412 frpr_icmp6(fin);
413 go = 0;
414 break;
415
416 case IPPROTO_GRE :
417 frpr_gre6(fin);
418 go = 0;
419 break;
420
421 case IPPROTO_HOPOPTS :
422 /*
423 * hop by hop ext header is only allowed
424 * right after IPv6 header.
425 */
426 if (hdrcount != 0) {
427 fin->fin_flx |= FI_BAD;
428 p = IPPROTO_NONE;
429 } else {
430 p = frpr_hopopts6(fin);
431 }
432 break;
433
434 case IPPROTO_DSTOPTS :
435 p = frpr_dstopts6(fin);
436 break;
437
438 case IPPROTO_ROUTING :
439 p = frpr_routing6(fin);
440 break;
441
442 case IPPROTO_AH :
443 p = frpr_ah6(fin);
444 break;
445
446 case IPPROTO_ESP :
447 frpr_esp6(fin);
448 go = 0;
449 break;
450
451 case IPPROTO_IPV6 :
452 for (i = 0; ip6exthdr[i].ol_bit != 0; i++)
453 if (ip6exthdr[i].ol_val == p) {
454 fin->fin_flx |= ip6exthdr[i].ol_bit;
455 break;
456 }
457 go = 0;
458 break;
459
460 case IPPROTO_NONE :
461 go = 0;
462 break;
463
464 case IPPROTO_FRAGMENT :
465 p = frpr_fragment6(fin);
466 if (fin->fin_off != 0) /* Not the first frag */
467 go = 0;
468 break;
469
470 default :
471 go = 0;
472 break;
473 }
474 hdrcount++;
475
476 /*
477 * It is important to note that at this point, for the
478 * extension headers (go != 0), the entire header may not have
479 * been pulled up when the code gets to this point. This is
480 * only done for "go != 0" because the other header handlers
481 * will all pullup their complete header. The other indicator
482 * of an incomplete packet is that this was just an extension
483 * header.
484 */
485 if ((go != 0) && (p != IPPROTO_NONE) &&
486 (frpr_pullup(fin, 0) == -1)) {
487 p = IPPROTO_NONE;
488 go = 0;
489 }
490 }
491 fi->fi_p = p;
492 }
493
494
495 /* ------------------------------------------------------------------------ */
496 /* Function: frpr_ipv6exthdr */
497 /* Returns: int - value of the next header or IPPROTO_NONE if error */
498 /* Parameters: fin(I) - pointer to packet information */
499 /* multiple(I) - flag indicating yes/no if multiple occurances */
500 /* of this extension header are allowed. */
501 /* proto(I) - protocol number for this extension header */
502 /* */
503 /* IPv6 Only */
504 /* This function expects to find an IPv6 extension header at fin_dp. */
505 /* There must be at least 8 bytes of data at fin_dp for there to be a valid */
506 /* extension header present. If a good one is found, fin_dp is advanced to */
507 /* point at the first piece of data after the extension header, fin_exthdr */
508 /* points to the start of the extension header and the "protocol" of the */
509 /* *NEXT* header is returned. */
510 /* ------------------------------------------------------------------------ */
frpr_ipv6exthdr(fin,multiple,proto)511 static INLINE int frpr_ipv6exthdr(fin, multiple, proto)
512 fr_info_t *fin;
513 int multiple, proto;
514 {
515 struct ip6_ext *hdr;
516 u_short shift;
517 int i;
518
519 fin->fin_flx |= FI_V6EXTHDR;
520
521 /* 8 is default length of extension hdr */
522 if ((fin->fin_dlen - 8) < 0) {
523 fin->fin_flx |= FI_SHORT;
524 return IPPROTO_NONE;
525 }
526
527 if (frpr_pullup(fin, 8) == -1)
528 return IPPROTO_NONE;
529
530 hdr = fin->fin_dp;
531 shift = 8 + (hdr->ip6e_len << 3);
532 if (shift > fin->fin_dlen) { /* Nasty extension header length? */
533 fin->fin_flx |= FI_BAD;
534 return IPPROTO_NONE;
535 }
536
537 for (i = 0; ip6exthdr[i].ol_bit != 0; i++)
538 if (ip6exthdr[i].ol_val == proto) {
539 /*
540 * Most IPv6 extension headers are only allowed once.
541 */
542 if ((multiple == 0) &&
543 ((fin->fin_optmsk & ip6exthdr[i].ol_bit) != 0))
544 fin->fin_flx |= FI_BAD;
545 else
546 fin->fin_optmsk |= ip6exthdr[i].ol_bit;
547 break;
548 }
549
550 fin->fin_dp = (char *)fin->fin_dp + shift;
551 fin->fin_dlen -= shift;
552
553 return hdr->ip6e_nxt;
554 }
555
556
557 /* ------------------------------------------------------------------------ */
558 /* Function: frpr_hopopts6 */
559 /* Returns: int - value of the next header or IPPROTO_NONE if error */
560 /* Parameters: fin(I) - pointer to packet information */
561 /* */
562 /* IPv6 Only */
563 /* This is function checks pending hop by hop options extension header */
564 /* ------------------------------------------------------------------------ */
frpr_hopopts6(fin)565 static INLINE int frpr_hopopts6(fin)
566 fr_info_t *fin;
567 {
568 return frpr_ipv6exthdr(fin, 0, IPPROTO_HOPOPTS);
569 }
570
571
572 /* ------------------------------------------------------------------------ */
573 /* Function: frpr_routing6 */
574 /* Returns: int - value of the next header or IPPROTO_NONE if error */
575 /* Parameters: fin(I) - pointer to packet information */
576 /* */
577 /* IPv6 Only */
578 /* This is function checks pending routing extension header */
579 /* ------------------------------------------------------------------------ */
frpr_routing6(fin)580 static INLINE int frpr_routing6(fin)
581 fr_info_t *fin;
582 {
583 struct ip6_ext *hdr;
584 int shift;
585
586 hdr = fin->fin_dp;
587 if (frpr_ipv6exthdr(fin, 0, IPPROTO_ROUTING) == IPPROTO_NONE)
588 return IPPROTO_NONE;
589
590 shift = 8 + (hdr->ip6e_len << 3);
591 /*
592 * Nasty extension header length?
593 */
594 if ((hdr->ip6e_len << 3) & 15) {
595 fin->fin_flx |= FI_BAD;
596 /*
597 * Compensate for the changes made in frpr_ipv6exthdr()
598 */
599 fin->fin_dlen += shift;
600 fin->fin_dp = (char *)fin->fin_dp - shift;
601 return IPPROTO_NONE;
602 }
603
604 return hdr->ip6e_nxt;
605 }
606
607
608 /* ------------------------------------------------------------------------ */
609 /* Function: frpr_fragment6 */
610 /* Returns: int - value of the next header or IPPROTO_NONE if error */
611 /* Parameters: fin(I) - pointer to packet information */
612 /* */
613 /* IPv6 Only */
614 /* Examine the IPv6 fragment header and extract fragment offset information.*/
615 /* */
616 /* We don't know where the transport layer header (or whatever is next is), */
617 /* as it could be behind destination options (amongst others). Because */
618 /* there is no fragment cache, there is no knowledge about whether or not an*/
619 /* upper layer header has been seen (or where it ends) and thus we are not */
620 /* able to continue processing beyond this header with any confidence. */
621 /* ------------------------------------------------------------------------ */
frpr_fragment6(fin)622 static INLINE int frpr_fragment6(fin)
623 fr_info_t *fin;
624 {
625 struct ip6_frag *frag;
626
627 fin->fin_flx |= FI_FRAG;
628
629 /*
630 * A fragmented IPv6 packet implies that there must be something
631 * else after the fragment.
632 */
633 if (frpr_ipv6exthdr(fin, 0, IPPROTO_FRAGMENT) == IPPROTO_NONE)
634 return IPPROTO_NONE;
635
636 frag = (struct ip6_frag *)((char *)fin->fin_dp - sizeof(*frag));
637
638 /*
639 * If this fragment isn't the last then the packet length must
640 * be a multiple of 8.
641 */
642 if ((frag->ip6f_offlg & IP6F_MORE_FRAG) != 0) {
643 fin->fin_flx |= FI_MOREFRAG;
644
645 if ((fin->fin_plen & 0x7) != 0)
646 fin->fin_flx |= FI_BAD;
647 }
648
649 fin->fin_id = frag->ip6f_ident;
650 fin->fin_off = ntohs(frag->ip6f_offlg & IP6F_OFF_MASK);
651 if (fin->fin_off != 0)
652 fin->fin_flx |= FI_FRAGBODY;
653
654 return frag->ip6f_nxt;
655 }
656
657
658 /* ------------------------------------------------------------------------ */
659 /* Function: frpr_dstopts6 */
660 /* Returns: int - value of the next header or IPPROTO_NONE if error */
661 /* Parameters: fin(I) - pointer to packet information */
662 /* nextheader(I) - stores next header value */
663 /* */
664 /* IPv6 Only */
665 /* This is function checks pending destination options extension header */
666 /* ------------------------------------------------------------------------ */
frpr_dstopts6(fin)667 static INLINE int frpr_dstopts6(fin)
668 fr_info_t *fin;
669 {
670 return frpr_ipv6exthdr(fin, 1, IPPROTO_DSTOPTS);
671 }
672
673
674 /* ------------------------------------------------------------------------ */
675 /* Function: frpr_icmp6 */
676 /* Returns: void */
677 /* Parameters: fin(I) - pointer to packet information */
678 /* */
679 /* IPv6 Only */
680 /* This routine is mainly concerned with determining the minimum valid size */
681 /* for an ICMPv6 packet. */
682 /* ------------------------------------------------------------------------ */
frpr_icmp6(fin)683 static INLINE void frpr_icmp6(fin)
684 fr_info_t *fin;
685 {
686 int minicmpsz = sizeof(struct icmp6_hdr);
687 struct icmp6_hdr *icmp6;
688
689 if (frpr_pullup(fin, ICMP6ERR_MINPKTLEN - sizeof(ip6_t)) == -1)
690 return;
691
692 if (fin->fin_dlen > 1) {
693 icmp6 = fin->fin_dp;
694
695 fin->fin_data[0] = *(u_short *)icmp6;
696
697 if ((icmp6->icmp6_type & ICMP6_INFOMSG_MASK) != 0)
698 fin->fin_flx |= FI_ICMPQUERY;
699
700 switch (icmp6->icmp6_type)
701 {
702 case ICMP6_ECHO_REPLY :
703 case ICMP6_ECHO_REQUEST :
704 if (fin->fin_dlen >= 6)
705 fin->fin_data[1] = icmp6->icmp6_id;
706 minicmpsz = ICMP6ERR_MINPKTLEN - sizeof(ip6_t);
707 break;
708 case ICMP6_DST_UNREACH :
709 case ICMP6_PACKET_TOO_BIG :
710 case ICMP6_TIME_EXCEEDED :
711 case ICMP6_PARAM_PROB :
712 if ((fin->fin_m != NULL) &&
713 (M_LEN(fin->fin_m) < fin->fin_plen)) {
714 if (fr_coalesce(fin) != 1)
715 return;
716 }
717 fin->fin_flx |= FI_ICMPERR;
718 minicmpsz = ICMP6ERR_IPICMPHLEN - sizeof(ip6_t);
719 break;
720 default :
721 break;
722 }
723 }
724
725 frpr_short6(fin, minicmpsz);
726 }
727
728
729 /* ------------------------------------------------------------------------ */
730 /* Function: frpr_udp6 */
731 /* Returns: void */
732 /* Parameters: fin(I) - pointer to packet information */
733 /* */
734 /* IPv6 Only */
735 /* Analyse the packet for IPv6/UDP properties. */
736 /* Is not expected to be called for fragmented packets. */
737 /* ------------------------------------------------------------------------ */
frpr_udp6(fin)738 static INLINE void frpr_udp6(fin)
739 fr_info_t *fin;
740 {
741
742 fr_checkv6sum(fin);
743
744 frpr_short6(fin, sizeof(struct udphdr));
745 if (frpr_pullup(fin, sizeof(struct udphdr)) == -1)
746 return;
747
748 frpr_udpcommon(fin);
749 }
750
751
752 /* ------------------------------------------------------------------------ */
753 /* Function: frpr_tcp6 */
754 /* Returns: void */
755 /* Parameters: fin(I) - pointer to packet information */
756 /* */
757 /* IPv6 Only */
758 /* Analyse the packet for IPv6/TCP properties. */
759 /* Is not expected to be called for fragmented packets. */
760 /* ------------------------------------------------------------------------ */
frpr_tcp6(fin)761 static INLINE void frpr_tcp6(fin)
762 fr_info_t *fin;
763 {
764
765 fr_checkv6sum(fin);
766
767 frpr_short6(fin, sizeof(struct tcphdr));
768 if (frpr_pullup(fin, sizeof(struct tcphdr)) == -1)
769 return;
770
771 frpr_tcpcommon(fin);
772 }
773
774
775 /* ------------------------------------------------------------------------ */
776 /* Function: frpr_esp6 */
777 /* Returns: void */
778 /* Parameters: fin(I) - pointer to packet information */
779 /* */
780 /* IPv6 Only */
781 /* Analyse the packet for ESP properties. */
782 /* The minimum length is taken to be the SPI (32bits) plus a tail (32bits) */
783 /* even though the newer ESP packets must also have a sequence number that */
784 /* is 32bits as well, it is not possible(?) to determine the version from a */
785 /* simple packet header. */
786 /* ------------------------------------------------------------------------ */
frpr_esp6(fin)787 static INLINE void frpr_esp6(fin)
788 fr_info_t *fin;
789 {
790 int i;
791 frpr_short6(fin, sizeof(grehdr_t));
792
793 (void) frpr_pullup(fin, 8);
794
795 for (i = 0; ip6exthdr[i].ol_bit != 0; i++)
796 if (ip6exthdr[i].ol_val == IPPROTO_ESP) {
797 fin->fin_optmsk |= ip6exthdr[i].ol_bit;
798 break;
799 }
800 }
801
802
803 /* ------------------------------------------------------------------------ */
804 /* Function: frpr_ah6 */
805 /* Returns: void */
806 /* Parameters: fin(I) - pointer to packet information */
807 /* */
808 /* IPv6 Only */
809 /* Analyse the packet for AH properties. */
810 /* The minimum length is taken to be the combination of all fields in the */
811 /* header being present and no authentication data (null algorithm used.) */
812 /* ------------------------------------------------------------------------ */
frpr_ah6(fin)813 static INLINE int frpr_ah6(fin)
814 fr_info_t *fin;
815 {
816 authhdr_t *ah;
817 int i, shift;
818
819 frpr_short6(fin, 12);
820
821 if (frpr_pullup(fin, sizeof(*ah)) == -1)
822 return IPPROTO_NONE;
823
824 for (i = 0; ip6exthdr[i].ol_bit != 0; i++)
825 if (ip6exthdr[i].ol_val == IPPROTO_AH) {
826 fin->fin_optmsk |= ip6exthdr[i].ol_bit;
827 break;
828 }
829
830 ah = (authhdr_t *)fin->fin_dp;
831
832 shift = (ah->ah_plen + 2) * 4;
833 fin->fin_dlen -= shift;
834 fin->fin_dp = (char*)fin->fin_dp + shift;
835
836 return ah->ah_next;
837 }
838
839
840 /* ------------------------------------------------------------------------ */
841 /* Function: frpr_gre6 */
842 /* Returns: void */
843 /* Parameters: fin(I) - pointer to packet information */
844 /* */
845 /* Analyse the packet for GRE properties. */
846 /* ------------------------------------------------------------------------ */
frpr_gre6(fin)847 static INLINE void frpr_gre6(fin)
848 fr_info_t *fin;
849 {
850 grehdr_t *gre;
851
852 frpr_short6(fin, sizeof(grehdr_t));
853
854 if (frpr_pullup(fin, sizeof(grehdr_t)) == -1)
855 return;
856
857 gre = fin->fin_dp;
858 if (GRE_REV(gre->gr_flags) == 1)
859 fin->fin_data[0] = gre->gr_call;
860 }
861 #endif /* USE_INET6 */
862
863
864 /* ------------------------------------------------------------------------ */
865 /* Function: frpr_pullup */
866 /* Returns: int - 0 == pullup succeeded, -1 == failure */
867 /* Parameters: fin(I) - pointer to packet information */
868 /* plen(I) - length (excluding L3 header) to pullup */
869 /* */
870 /* Short inline function to cut down on code duplication to perform a call */
871 /* to fr_pullup to ensure there is the required amount of data, */
872 /* consecutively in the packet buffer. */
873 /* ------------------------------------------------------------------------ */
frpr_pullup(fin,plen)874 static INLINE int frpr_pullup(fin, plen)
875 fr_info_t *fin;
876 int plen;
877 {
878 #if defined(_KERNEL)
879 if (fin->fin_m != NULL) {
880 int ipoff;
881
882 ipoff = (char *)fin->fin_ip - MTOD(fin->fin_m, char *);
883
884 if (fin->fin_dp != NULL)
885 plen += (char *)fin->fin_dp -
886 ((char *)fin->fin_ip + fin->fin_hlen);
887 plen += fin->fin_hlen;
888 /*
889 * We don't do 'plen += ipoff;' here. The fr_pullup() will
890 * do it for us.
891 */
892 if (M_LEN(fin->fin_m) < plen + ipoff) {
893 if (fr_pullup(fin->fin_m, fin, plen) == NULL)
894 return -1;
895 }
896 }
897 #endif
898 return 0;
899 }
900
901
902 /* ------------------------------------------------------------------------ */
903 /* Function: frpr_short */
904 /* Returns: void */
905 /* Parameters: fin(I) - pointer to packet information */
906 /* xmin(I) - minimum header size */
907 /* */
908 /* Check if a packet is "short" as defined by xmin. The rule we are */
909 /* applying here is that the packet must not be fragmented within the layer */
910 /* 4 header. That is, it must not be a fragment that has its offset set to */
911 /* start within the layer 4 header (hdrmin) or if it is at offset 0, the */
912 /* entire layer 4 header must be present (min). */
913 /* ------------------------------------------------------------------------ */
frpr_short(fin,xmin)914 static INLINE void frpr_short(fin, xmin)
915 fr_info_t *fin;
916 int xmin;
917 {
918
919 if (fin->fin_off == 0) {
920 if (fin->fin_dlen < xmin)
921 fin->fin_flx |= FI_SHORT;
922 } else if (fin->fin_off < xmin) {
923 fin->fin_flx |= FI_SHORT;
924 }
925 }
926
927
928 /* ------------------------------------------------------------------------ */
929 /* Function: frpr_icmp */
930 /* Returns: void */
931 /* Parameters: fin(I) - pointer to packet information */
932 /* */
933 /* IPv4 Only */
934 /* Do a sanity check on the packet for ICMP (v4). In nearly all cases, */
935 /* except extrememly bad packets, both type and code will be present. */
936 /* The expected minimum size of an ICMP packet is very much dependent on */
937 /* the type of it. */
938 /* */
939 /* XXX - other ICMP sanity checks? */
940 /* ------------------------------------------------------------------------ */
frpr_icmp(fin)941 static INLINE void frpr_icmp(fin)
942 fr_info_t *fin;
943 {
944 int minicmpsz = sizeof(struct icmp);
945 icmphdr_t *icmp;
946 ip_t *oip;
947 ipf_stack_t *ifs = fin->fin_ifs;
948
949 if (fin->fin_off != 0) {
950 frpr_short(fin, ICMPERR_ICMPHLEN);
951 return;
952 }
953
954 if (frpr_pullup(fin, ICMPERR_ICMPHLEN) == -1)
955 return;
956
957 fr_checkv4sum(fin);
958
959 /*
960 * This is a right place to set icmp pointer, since the memory
961 * referenced by fin_dp could get reallocated. The code down below can
962 * rely on fact icmp variable always points to ICMP header.
963 */
964 icmp = fin->fin_dp;
965 fin->fin_data[0] = *(u_short *)icmp;
966 fin->fin_data[1] = icmp->icmp_id;
967
968 switch (icmp->icmp_type)
969 {
970 case ICMP_ECHOREPLY :
971 case ICMP_ECHO :
972 /* Router discovery messaes - RFC 1256 */
973 case ICMP_ROUTERADVERT :
974 case ICMP_ROUTERSOLICIT :
975 minicmpsz = ICMP_MINLEN;
976 break;
977 /*
978 * type(1) + code(1) + cksum(2) + id(2) seq(2) +
979 * 3 * timestamp(3 * 4)
980 */
981 case ICMP_TSTAMP :
982 case ICMP_TSTAMPREPLY :
983 minicmpsz = 20;
984 break;
985 /*
986 * type(1) + code(1) + cksum(2) + id(2) seq(2) +
987 * mask(4)
988 */
989 case ICMP_MASKREQ :
990 case ICMP_MASKREPLY :
991 minicmpsz = 12;
992 break;
993 /*
994 * type(1) + code(1) + cksum(2) + id(2) seq(2) + ip(20+)
995 */
996 case ICMP_UNREACH :
997 if (icmp->icmp_code == ICMP_UNREACH_NEEDFRAG) {
998 if (icmp->icmp_nextmtu < ifs->ifs_fr_icmpminfragmtu)
999 fin->fin_flx |= FI_BAD;
1000 }
1001 /* FALLTHRU */
1002 case ICMP_SOURCEQUENCH :
1003 case ICMP_REDIRECT :
1004 case ICMP_TIMXCEED :
1005 case ICMP_PARAMPROB :
1006 fin->fin_flx |= FI_ICMPERR;
1007 if (fr_coalesce(fin) != 1)
1008 return;
1009 /*
1010 * ICMP error packets should not be generated for IP
1011 * packets that are a fragment that isn't the first
1012 * fragment.
1013 */
1014 oip = (ip_t *)((char *)fin->fin_dp + ICMPERR_ICMPHLEN);
1015 if ((ntohs(oip->ip_off) & IP_OFFMASK) != 0)
1016 fin->fin_flx |= FI_BAD;
1017 break;
1018 default :
1019 break;
1020 }
1021
1022 frpr_short(fin, minicmpsz);
1023 }
1024
1025
1026 /* ------------------------------------------------------------------------ */
1027 /* Function: frpr_tcpcommon */
1028 /* Returns: void */
1029 /* Parameters: fin(I) - pointer to packet information */
1030 /* */
1031 /* TCP header sanity checking. Look for bad combinations of TCP flags, */
1032 /* and make some checks with how they interact with other fields. */
1033 /* If compiled with IPFILTER_CKSUM, check to see if the TCP checksum is */
1034 /* valid and mark the packet as bad if not. */
1035 /* ------------------------------------------------------------------------ */
frpr_tcpcommon(fin)1036 static INLINE void frpr_tcpcommon(fin)
1037 fr_info_t *fin;
1038 {
1039 int flags, tlen;
1040 tcphdr_t *tcp;
1041
1042 fin->fin_flx |= FI_TCPUDP;
1043 if (fin->fin_off != 0)
1044 return;
1045
1046 if (frpr_pullup(fin, sizeof(*tcp)) == -1)
1047 return;
1048 tcp = fin->fin_dp;
1049
1050 if (fin->fin_dlen > 3) {
1051 fin->fin_sport = ntohs(tcp->th_sport);
1052 fin->fin_dport = ntohs(tcp->th_dport);
1053 }
1054
1055 if ((fin->fin_flx & FI_SHORT) != 0)
1056 return;
1057
1058 /*
1059 * Use of the TCP data offset *must* result in a value that is at
1060 * least the same size as the TCP header.
1061 */
1062 tlen = TCP_OFF(tcp) << 2;
1063 if (tlen < sizeof(tcphdr_t)) {
1064 fin->fin_flx |= FI_BAD;
1065 return;
1066 }
1067
1068 flags = tcp->th_flags;
1069 fin->fin_tcpf = tcp->th_flags;
1070
1071 /*
1072 * If the urgent flag is set, then the urgent pointer must
1073 * also be set and vice versa. Good TCP packets do not have
1074 * just one of these set.
1075 */
1076 if ((flags & TH_URG) != 0 && (tcp->th_urp == 0)) {
1077 fin->fin_flx |= FI_BAD;
1078 } else if ((flags & TH_URG) == 0 && (tcp->th_urp != 0)) {
1079 /* Ignore this case, it shows up in "real" traffic with */
1080 /* bogus values in the urgent pointer field. */
1081 flags = flags; /* LINT */
1082 } else if (((flags & (TH_SYN|TH_FIN)) != 0) &&
1083 ((flags & (TH_RST|TH_ACK)) == TH_RST)) {
1084 /* TH_FIN|TH_RST|TH_ACK seems to appear "naturally" */
1085 fin->fin_flx |= FI_BAD;
1086 } else if (!(flags & TH_ACK)) {
1087 /*
1088 * If the ack bit isn't set, then either the SYN or
1089 * RST bit must be set. If the SYN bit is set, then
1090 * we expect the ACK field to be 0. If the ACK is
1091 * not set and if URG, PSH or FIN are set, consdier
1092 * that to indicate a bad TCP packet.
1093 */
1094 if ((flags == TH_SYN) && (tcp->th_ack != 0)) {
1095 /*
1096 * Cisco PIX sets the ACK field to a random value.
1097 * In light of this, do not set FI_BAD until a patch
1098 * is available from Cisco to ensure that
1099 * interoperability between existing systems is
1100 * achieved.
1101 */
1102 /*fin->fin_flx |= FI_BAD*/;
1103 flags = flags; /* LINT */
1104 } else if (!(flags & (TH_RST|TH_SYN))) {
1105 fin->fin_flx |= FI_BAD;
1106 } else if ((flags & (TH_URG|TH_PUSH|TH_FIN)) != 0) {
1107 fin->fin_flx |= FI_BAD;
1108 }
1109 }
1110
1111 /*
1112 * At this point, it's not exactly clear what is to be gained by
1113 * marking up which TCP options are and are not present. The one we
1114 * are most interested in is the TCP window scale. This is only in
1115 * a SYN packet [RFC1323] so we don't need this here...?
1116 * Now if we were to analyse the header for passive fingerprinting,
1117 * then that might add some weight to adding this...
1118 */
1119 if (tlen == sizeof(tcphdr_t))
1120 return;
1121
1122 if (frpr_pullup(fin, tlen) == -1)
1123 return;
1124
1125 #if 0
1126 ip = fin->fin_ip;
1127 s = (u_char *)(tcp + 1);
1128 off = IP_HL(ip) << 2;
1129 # ifdef _KERNEL
1130 if (fin->fin_mp != NULL) {
1131 mb_t *m = *fin->fin_mp;
1132
1133 if (off + tlen > M_LEN(m))
1134 return;
1135 }
1136 # endif
1137 for (tlen -= (int)sizeof(*tcp); tlen > 0; ) {
1138 opt = *s;
1139 if (opt == '\0')
1140 break;
1141 else if (opt == TCPOPT_NOP)
1142 ol = 1;
1143 else {
1144 if (tlen < 2)
1145 break;
1146 ol = (int)*(s + 1);
1147 if (ol < 2 || ol > tlen)
1148 break;
1149 }
1150
1151 for (i = 9, mv = 4; mv >= 0; ) {
1152 op = ipopts + i;
1153 if (opt == (u_char)op->ol_val) {
1154 optmsk |= op->ol_bit;
1155 break;
1156 }
1157 }
1158 tlen -= ol;
1159 s += ol;
1160 }
1161 #endif /* 0 */
1162 }
1163
1164
1165
1166 /* ------------------------------------------------------------------------ */
1167 /* Function: frpr_udpcommon */
1168 /* Returns: void */
1169 /* Parameters: fin(I) - pointer to packet information */
1170 /* */
1171 /* Extract the UDP source and destination ports, if present. If compiled */
1172 /* with IPFILTER_CKSUM, check to see if the UDP checksum is valid. */
1173 /* ------------------------------------------------------------------------ */
frpr_udpcommon(fin)1174 static INLINE void frpr_udpcommon(fin)
1175 fr_info_t *fin;
1176 {
1177 udphdr_t *udp;
1178
1179 fin->fin_flx |= FI_TCPUDP;
1180
1181 if (!fin->fin_off && (fin->fin_dlen > 3)) {
1182 if (frpr_pullup(fin, sizeof(*udp)) == -1) {
1183 fin->fin_flx |= FI_SHORT;
1184 return;
1185 }
1186
1187 udp = fin->fin_dp;
1188
1189 fin->fin_sport = ntohs(udp->uh_sport);
1190 fin->fin_dport = ntohs(udp->uh_dport);
1191 }
1192 }
1193
1194
1195 /* ------------------------------------------------------------------------ */
1196 /* Function: frpr_tcp */
1197 /* Returns: void */
1198 /* Parameters: fin(I) - pointer to packet information */
1199 /* */
1200 /* IPv4 Only */
1201 /* Analyse the packet for IPv4/TCP properties. */
1202 /* ------------------------------------------------------------------------ */
frpr_tcp(fin)1203 static INLINE void frpr_tcp(fin)
1204 fr_info_t *fin;
1205 {
1206
1207 fr_checkv4sum(fin);
1208
1209 frpr_short(fin, sizeof(tcphdr_t));
1210
1211 frpr_tcpcommon(fin);
1212 }
1213
1214
1215 /* ------------------------------------------------------------------------ */
1216 /* Function: frpr_udp */
1217 /* Returns: void */
1218 /* Parameters: fin(I) - pointer to packet information */
1219 /* */
1220 /* IPv4 Only */
1221 /* Analyse the packet for IPv4/UDP properties. */
1222 /* ------------------------------------------------------------------------ */
frpr_udp(fin)1223 static INLINE void frpr_udp(fin)
1224 fr_info_t *fin;
1225 {
1226
1227 fr_checkv4sum(fin);
1228
1229 frpr_short(fin, sizeof(udphdr_t));
1230
1231 frpr_udpcommon(fin);
1232 }
1233
1234
1235 /* ------------------------------------------------------------------------ */
1236 /* Function: frpr_esp */
1237 /* Returns: void */
1238 /* Parameters: fin(I) - pointer to packet information */
1239 /* */
1240 /* Analyse the packet for ESP properties. */
1241 /* The minimum length is taken to be the SPI (32bits) plus a tail (32bits) */
1242 /* even though the newer ESP packets must also have a sequence number that */
1243 /* is 32bits as well, it is not possible(?) to determine the version from a */
1244 /* simple packet header. */
1245 /* ------------------------------------------------------------------------ */
frpr_esp(fin)1246 static INLINE void frpr_esp(fin)
1247 fr_info_t *fin;
1248 {
1249 if ((fin->fin_off == 0) && (frpr_pullup(fin, 8) == -1))
1250 return;
1251
1252 frpr_short(fin, 8);
1253 }
1254
1255
1256 /* ------------------------------------------------------------------------ */
1257 /* Function: frpr_ah */
1258 /* Returns: void */
1259 /* Parameters: fin(I) - pointer to packet information */
1260 /* */
1261 /* Analyse the packet for AH properties. */
1262 /* The minimum length is taken to be the combination of all fields in the */
1263 /* header being present and no authentication data (null algorithm used.) */
1264 /* ------------------------------------------------------------------------ */
frpr_ah(fin)1265 static INLINE void frpr_ah(fin)
1266 fr_info_t *fin;
1267 {
1268 authhdr_t *ah;
1269 int len;
1270
1271 if ((fin->fin_off == 0) && (frpr_pullup(fin, sizeof(*ah)) == -1))
1272 return;
1273
1274 ah = (authhdr_t *)fin->fin_dp;
1275
1276 len = (ah->ah_plen + 2) << 2;
1277 frpr_short(fin, len);
1278 }
1279
1280
1281 /* ------------------------------------------------------------------------ */
1282 /* Function: frpr_gre */
1283 /* Returns: void */
1284 /* Parameters: fin(I) - pointer to packet information */
1285 /* */
1286 /* Analyse the packet for GRE properties. */
1287 /* ------------------------------------------------------------------------ */
frpr_gre(fin)1288 static INLINE void frpr_gre(fin)
1289 fr_info_t *fin;
1290 {
1291 grehdr_t *gre;
1292
1293 if ((fin->fin_off == 0) && (frpr_pullup(fin, sizeof(grehdr_t)) == -1))
1294 return;
1295
1296 frpr_short(fin, sizeof(grehdr_t));
1297
1298 if (fin->fin_off == 0) {
1299 gre = fin->fin_dp;
1300 if (GRE_REV(gre->gr_flags) == 1)
1301 fin->fin_data[0] = gre->gr_call;
1302 }
1303 }
1304
1305
1306 /* ------------------------------------------------------------------------ */
1307 /* Function: frpr_ipv4hdr */
1308 /* Returns: void */
1309 /* Parameters: fin(I) - pointer to packet information */
1310 /* */
1311 /* IPv4 Only */
1312 /* Analyze the IPv4 header and set fields in the fr_info_t structure. */
1313 /* Check all options present and flag their presence if any exist. */
1314 /* ------------------------------------------------------------------------ */
frpr_ipv4hdr(fin)1315 static INLINE void frpr_ipv4hdr(fin)
1316 fr_info_t *fin;
1317 {
1318 u_short optmsk = 0, secmsk = 0, auth = 0;
1319 int hlen, ol, mv, p, i;
1320 const struct optlist *op;
1321 u_char *s, opt;
1322 u_short off;
1323 fr_ip_t *fi;
1324 ip_t *ip;
1325
1326 fi = &fin->fin_fi;
1327 hlen = fin->fin_hlen;
1328
1329 ip = fin->fin_ip;
1330 p = ip->ip_p;
1331 fi->fi_p = p;
1332 fi->fi_tos = ip->ip_tos;
1333 fin->fin_id = ip->ip_id;
1334 off = ip->ip_off;
1335
1336 /* Get both TTL and protocol */
1337 fi->fi_p = ip->ip_p;
1338 fi->fi_ttl = ip->ip_ttl;
1339 #if 0
1340 (*(((u_short *)fi) + 1)) = (*(((u_short *)ip) + 4));
1341 #endif
1342
1343 /* Zero out bits not used in IPv6 address */
1344 fi->fi_src.i6[1] = 0;
1345 fi->fi_src.i6[2] = 0;
1346 fi->fi_src.i6[3] = 0;
1347 fi->fi_dst.i6[1] = 0;
1348 fi->fi_dst.i6[2] = 0;
1349 fi->fi_dst.i6[3] = 0;
1350
1351 fi->fi_saddr = ip->ip_src.s_addr;
1352 fi->fi_daddr = ip->ip_dst.s_addr;
1353
1354 /*
1355 * set packet attribute flags based on the offset and
1356 * calculate the byte offset that it represents.
1357 */
1358 off &= IP_MF|IP_OFFMASK;
1359 if (off != 0) {
1360 int morefrag = off & IP_MF;
1361
1362 fi->fi_flx |= FI_FRAG;
1363 if (morefrag)
1364 fi->fi_flx |= FI_MOREFRAG;
1365 off &= IP_OFFMASK;
1366 if (off != 0) {
1367 fin->fin_flx |= FI_FRAGBODY;
1368 off <<= 3;
1369 if ((off + fin->fin_dlen > 65535) ||
1370 (fin->fin_dlen == 0) ||
1371 ((morefrag != 0) && ((fin->fin_dlen & 7) != 0))) {
1372 /*
1373 * The length of the packet, starting at its
1374 * offset cannot exceed 65535 (0xffff) as the
1375 * length of an IP packet is only 16 bits.
1376 *
1377 * Any fragment that isn't the last fragment
1378 * must have a length greater than 0 and it
1379 * must be an even multiple of 8.
1380 */
1381 fi->fi_flx |= FI_BAD;
1382 }
1383 }
1384 }
1385 fin->fin_off = off;
1386
1387 /*
1388 * Call per-protocol setup and checking
1389 */
1390 switch (p)
1391 {
1392 case IPPROTO_UDP :
1393 frpr_udp(fin);
1394 break;
1395 case IPPROTO_TCP :
1396 frpr_tcp(fin);
1397 break;
1398 case IPPROTO_ICMP :
1399 frpr_icmp(fin);
1400 break;
1401 case IPPROTO_AH :
1402 frpr_ah(fin);
1403 break;
1404 case IPPROTO_ESP :
1405 frpr_esp(fin);
1406 break;
1407 case IPPROTO_GRE :
1408 frpr_gre(fin);
1409 break;
1410 }
1411
1412 ip = fin->fin_ip;
1413 if (ip == NULL)
1414 return;
1415
1416 /*
1417 * If it is a standard IP header (no options), set the flag fields
1418 * which relate to options to 0.
1419 */
1420 if (hlen == sizeof(*ip)) {
1421 fi->fi_optmsk = 0;
1422 fi->fi_secmsk = 0;
1423 fi->fi_auth = 0;
1424 return;
1425 }
1426
1427 /*
1428 * So the IP header has some IP options attached. Walk the entire
1429 * list of options present with this packet and set flags to indicate
1430 * which ones are here and which ones are not. For the somewhat out
1431 * of date and obscure security classification options, set a flag to
1432 * represent which classification is present.
1433 */
1434 fi->fi_flx |= FI_OPTIONS;
1435
1436 for (s = (u_char *)(ip + 1), hlen -= (int)sizeof(*ip); hlen > 0; ) {
1437 opt = *s;
1438 if (opt == '\0')
1439 break;
1440 else if (opt == IPOPT_NOP)
1441 ol = 1;
1442 else {
1443 if (hlen < 2)
1444 break;
1445 ol = (int)*(s + 1);
1446 if (ol < 2 || ol > hlen)
1447 break;
1448 }
1449 for (i = 9, mv = 4; mv >= 0; ) {
1450 op = ipopts + i;
1451 if ((opt == (u_char)op->ol_val) && (ol > 4)) {
1452 optmsk |= op->ol_bit;
1453 if (opt == IPOPT_SECURITY) {
1454 const struct optlist *sp;
1455 u_char sec;
1456 int j, m;
1457
1458 sec = *(s + 2); /* classification */
1459 for (j = 3, m = 2; m >= 0; ) {
1460 sp = secopt + j;
1461 if (sec == sp->ol_val) {
1462 secmsk |= sp->ol_bit;
1463 auth = *(s + 3);
1464 auth *= 256;
1465 auth += *(s + 4);
1466 break;
1467 }
1468 if (sec < sp->ol_val)
1469 j -= m;
1470 else
1471 j += m;
1472 m--;
1473 }
1474 }
1475 break;
1476 }
1477 if (opt < op->ol_val)
1478 i -= mv;
1479 else
1480 i += mv;
1481 mv--;
1482 }
1483 hlen -= ol;
1484 s += ol;
1485 }
1486
1487 /*
1488 *
1489 */
1490 if (auth && !(auth & 0x0100))
1491 auth &= 0xff00;
1492 fi->fi_optmsk = optmsk;
1493 fi->fi_secmsk = secmsk;
1494 fi->fi_auth = auth;
1495 }
1496
1497
1498 /* ------------------------------------------------------------------------ */
1499 /* Function: fr_makefrip */
1500 /* Returns: int - 1 == hdr checking error, 0 == OK */
1501 /* Parameters: hlen(I) - length of IP packet header */
1502 /* ip(I) - pointer to the IP header */
1503 /* fin(IO) - pointer to packet information */
1504 /* */
1505 /* Compact the IP header into a structure which contains just the info. */
1506 /* which is useful for comparing IP headers with and store this information */
1507 /* in the fr_info_t structure pointer to by fin. At present, it is assumed */
1508 /* this function will be called with either an IPv4 or IPv6 packet. */
1509 /* ------------------------------------------------------------------------ */
fr_makefrip(hlen,ip,fin)1510 int fr_makefrip(hlen, ip, fin)
1511 int hlen;
1512 ip_t *ip;
1513 fr_info_t *fin;
1514 {
1515 int v;
1516
1517 fin->fin_depth = 0;
1518 fin->fin_hlen = (u_short)hlen;
1519 fin->fin_ip = ip;
1520 fin->fin_rule = 0xffffffff;
1521 fin->fin_group[0] = -1;
1522 fin->fin_group[1] = '\0';
1523 fin->fin_dlen = fin->fin_plen - hlen;
1524 fin->fin_dp = (char *)ip + hlen;
1525
1526 v = fin->fin_v;
1527 if (v == 4)
1528 frpr_ipv4hdr(fin);
1529 #ifdef USE_INET6
1530 else if (v == 6)
1531 frpr_ipv6hdr(fin);
1532 #endif
1533 if (fin->fin_ip == NULL)
1534 return -1;
1535 return 0;
1536 }
1537
1538
1539 /* ------------------------------------------------------------------------ */
1540 /* Function: fr_portcheck */
1541 /* Returns: int - 1 == port matched, 0 == port match failed */
1542 /* Parameters: frp(I) - pointer to port check `expression' */
1543 /* pop(I) - pointer to port number to evaluate */
1544 /* */
1545 /* Perform a comparison of a port number against some other(s), using a */
1546 /* structure with compare information stored in it. */
1547 /* ------------------------------------------------------------------------ */
fr_portcheck(frp,pop)1548 static INLINE int fr_portcheck(frp, pop)
1549 frpcmp_t *frp;
1550 u_short *pop;
1551 {
1552 u_short tup, po;
1553 int err = 1;
1554
1555 tup = *pop;
1556 po = frp->frp_port;
1557
1558 /*
1559 * Do opposite test to that required and continue if that succeeds.
1560 */
1561 switch (frp->frp_cmp)
1562 {
1563 case FR_EQUAL :
1564 if (tup != po) /* EQUAL */
1565 err = 0;
1566 break;
1567 case FR_NEQUAL :
1568 if (tup == po) /* NOTEQUAL */
1569 err = 0;
1570 break;
1571 case FR_LESST :
1572 if (tup >= po) /* LESSTHAN */
1573 err = 0;
1574 break;
1575 case FR_GREATERT :
1576 if (tup <= po) /* GREATERTHAN */
1577 err = 0;
1578 break;
1579 case FR_LESSTE :
1580 if (tup > po) /* LT or EQ */
1581 err = 0;
1582 break;
1583 case FR_GREATERTE :
1584 if (tup < po) /* GT or EQ */
1585 err = 0;
1586 break;
1587 case FR_OUTRANGE :
1588 if (tup >= po && tup <= frp->frp_top) /* Out of range */
1589 err = 0;
1590 break;
1591 case FR_INRANGE :
1592 if (tup <= po || tup >= frp->frp_top) /* In range */
1593 err = 0;
1594 break;
1595 case FR_INCRANGE :
1596 if (tup < po || tup > frp->frp_top) /* Inclusive range */
1597 err = 0;
1598 break;
1599 default :
1600 break;
1601 }
1602 return err;
1603 }
1604
1605
1606 /* ------------------------------------------------------------------------ */
1607 /* Function: fr_tcpudpchk */
1608 /* Returns: int - 1 == protocol matched, 0 == check failed */
1609 /* Parameters: fin(I) - pointer to packet information */
1610 /* ft(I) - pointer to structure with comparison data */
1611 /* */
1612 /* Compares the current pcket (assuming it is TCP/UDP) information with a */
1613 /* structure containing information that we want to match against. */
1614 /* ------------------------------------------------------------------------ */
fr_tcpudpchk(fin,ft)1615 int fr_tcpudpchk(fin, ft)
1616 fr_info_t *fin;
1617 frtuc_t *ft;
1618 {
1619 int err = 1;
1620
1621 /*
1622 * Both ports should *always* be in the first fragment.
1623 * So far, I cannot find any cases where they can not be.
1624 *
1625 * compare destination ports
1626 */
1627 if (ft->ftu_dcmp)
1628 err = fr_portcheck(&ft->ftu_dst, &fin->fin_dport);
1629
1630 /*
1631 * compare source ports
1632 */
1633 if (err && ft->ftu_scmp)
1634 err = fr_portcheck(&ft->ftu_src, &fin->fin_sport);
1635
1636 /*
1637 * If we don't have all the TCP/UDP header, then how can we
1638 * expect to do any sort of match on it ? If we were looking for
1639 * TCP flags, then NO match. If not, then match (which should
1640 * satisfy the "short" class too).
1641 */
1642 if (err && (fin->fin_p == IPPROTO_TCP)) {
1643 if (fin->fin_flx & FI_SHORT)
1644 return !(ft->ftu_tcpf | ft->ftu_tcpfm);
1645 /*
1646 * Match the flags ? If not, abort this match.
1647 */
1648 if (ft->ftu_tcpfm &&
1649 ft->ftu_tcpf != (fin->fin_tcpf & ft->ftu_tcpfm)) {
1650 FR_DEBUG(("f. %#x & %#x != %#x\n", fin->fin_tcpf,
1651 ft->ftu_tcpfm, ft->ftu_tcpf));
1652 err = 0;
1653 }
1654 }
1655 return err;
1656 }
1657
1658
1659 /* ------------------------------------------------------------------------ */
1660 /* Function: fr_ipfcheck */
1661 /* Returns: int - 0 == match, 1 == no match */
1662 /* Parameters: fin(I) - pointer to packet information */
1663 /* fr(I) - pointer to filter rule */
1664 /* portcmp(I) - flag indicating whether to attempt matching on */
1665 /* TCP/UDP port data. */
1666 /* */
1667 /* Check to see if a packet matches an IPFilter rule. Checks of addresses, */
1668 /* port numbers, etc, for "standard" IPFilter rules are all orchestrated in */
1669 /* this function. */
1670 /* ------------------------------------------------------------------------ */
fr_ipfcheck(fin,fr,portcmp)1671 static INLINE int fr_ipfcheck(fin, fr, portcmp)
1672 fr_info_t *fin;
1673 frentry_t *fr;
1674 int portcmp;
1675 {
1676 u_32_t *ld, *lm, *lip;
1677 fripf_t *fri;
1678 fr_ip_t *fi;
1679 int i;
1680 ipf_stack_t *ifs = fin->fin_ifs;
1681
1682 fi = &fin->fin_fi;
1683 fri = fr->fr_ipf;
1684 lip = (u_32_t *)fi;
1685 lm = (u_32_t *)&fri->fri_mip;
1686 ld = (u_32_t *)&fri->fri_ip;
1687
1688 /*
1689 * first 32 bits to check coversion:
1690 * IP version, TOS, TTL, protocol
1691 */
1692 i = ((*lip & *lm) != *ld);
1693 FR_DEBUG(("0. %#08x & %#08x != %#08x\n",
1694 *lip, *lm, *ld));
1695 if (i)
1696 return 1;
1697
1698 /*
1699 * Next 32 bits is a constructed bitmask indicating which IP options
1700 * are present (if any) in this packet.
1701 */
1702 lip++, lm++, ld++;
1703 i |= ((*lip & *lm) != *ld);
1704 FR_DEBUG(("1. %#08x & %#08x != %#08x\n",
1705 *lip, *lm, *ld));
1706 if (i)
1707 return 1;
1708
1709 lip++, lm++, ld++;
1710 /*
1711 * Unrolled loops (4 each, for 32 bits) for address checks.
1712 */
1713 /*
1714 * Check the source address.
1715 */
1716 #ifdef IPFILTER_LOOKUP
1717 if (fr->fr_satype == FRI_LOOKUP) {
1718 fin->fin_flx |= FI_DONTCACHE;
1719 i = (*fr->fr_srcfunc)(fr->fr_srcptr, fi->fi_v, lip, fin, ifs);
1720 if (i == -1)
1721 return 1;
1722 lip += 3;
1723 lm += 3;
1724 ld += 3;
1725 } else {
1726 #endif
1727 i = ((*lip & *lm) != *ld);
1728 FR_DEBUG(("2a. %#08x & %#08x != %#08x\n",
1729 *lip, *lm, *ld));
1730 if (fi->fi_v == 6) {
1731 lip++, lm++, ld++;
1732 i |= ((*lip & *lm) != *ld);
1733 FR_DEBUG(("2b. %#08x & %#08x != %#08x\n",
1734 *lip, *lm, *ld));
1735 lip++, lm++, ld++;
1736 i |= ((*lip & *lm) != *ld);
1737 FR_DEBUG(("2c. %#08x & %#08x != %#08x\n",
1738 *lip, *lm, *ld));
1739 lip++, lm++, ld++;
1740 i |= ((*lip & *lm) != *ld);
1741 FR_DEBUG(("2d. %#08x & %#08x != %#08x\n",
1742 *lip, *lm, *ld));
1743 } else {
1744 lip += 3;
1745 lm += 3;
1746 ld += 3;
1747 }
1748 #ifdef IPFILTER_LOOKUP
1749 }
1750 #endif
1751 i ^= (fr->fr_flags & FR_NOTSRCIP) >> 6;
1752 if (i)
1753 return 1;
1754
1755 /*
1756 * Check the destination address.
1757 */
1758 lip++, lm++, ld++;
1759 #ifdef IPFILTER_LOOKUP
1760 if (fr->fr_datype == FRI_LOOKUP) {
1761 fin->fin_flx |= FI_DONTCACHE;
1762 i = (*fr->fr_dstfunc)(fr->fr_dstptr, fi->fi_v, lip, fin, ifs);
1763 if (i == -1)
1764 return 1;
1765 lip += 3;
1766 lm += 3;
1767 ld += 3;
1768 } else {
1769 #endif
1770 i = ((*lip & *lm) != *ld);
1771 FR_DEBUG(("3a. %#08x & %#08x != %#08x\n",
1772 *lip, *lm, *ld));
1773 if (fi->fi_v == 6) {
1774 lip++, lm++, ld++;
1775 i |= ((*lip & *lm) != *ld);
1776 FR_DEBUG(("3b. %#08x & %#08x != %#08x\n",
1777 *lip, *lm, *ld));
1778 lip++, lm++, ld++;
1779 i |= ((*lip & *lm) != *ld);
1780 FR_DEBUG(("3c. %#08x & %#08x != %#08x\n",
1781 *lip, *lm, *ld));
1782 lip++, lm++, ld++;
1783 i |= ((*lip & *lm) != *ld);
1784 FR_DEBUG(("3d. %#08x & %#08x != %#08x\n",
1785 *lip, *lm, *ld));
1786 } else {
1787 lip += 3;
1788 lm += 3;
1789 ld += 3;
1790 }
1791 #ifdef IPFILTER_LOOKUP
1792 }
1793 #endif
1794 i ^= (fr->fr_flags & FR_NOTDSTIP) >> 7;
1795 if (i)
1796 return 1;
1797 /*
1798 * IP addresses matched. The next 32bits contains:
1799 * mast of old IP header security & authentication bits.
1800 */
1801 lip++, lm++, ld++;
1802 i |= ((*lip & *lm) != *ld);
1803 FR_DEBUG(("4. %#08x & %#08x != %#08x\n",
1804 *lip, *lm, *ld));
1805
1806 /*
1807 * Next we have 32 bits of packet flags.
1808 */
1809 lip++, lm++, ld++;
1810 i |= ((*lip & *lm) != *ld);
1811 FR_DEBUG(("5. %#08x & %#08x != %#08x\n",
1812 *lip, *lm, *ld));
1813
1814 if (i == 0) {
1815 /*
1816 * If a fragment, then only the first has what we're
1817 * looking for here...
1818 */
1819 if (portcmp) {
1820 if (!fr_tcpudpchk(fin, &fr->fr_tuc))
1821 i = 1;
1822 } else {
1823 if (fr->fr_dcmp || fr->fr_scmp ||
1824 fr->fr_tcpf || fr->fr_tcpfm)
1825 i = 1;
1826 if (fr->fr_icmpm || fr->fr_icmp) {
1827 if (((fi->fi_p != IPPROTO_ICMP) &&
1828 (fi->fi_p != IPPROTO_ICMPV6)) ||
1829 fin->fin_off || (fin->fin_dlen < 2))
1830 i = 1;
1831 else if ((fin->fin_data[0] & fr->fr_icmpm) !=
1832 fr->fr_icmp) {
1833 FR_DEBUG(("i. %#x & %#x != %#x\n",
1834 fin->fin_data[0],
1835 fr->fr_icmpm, fr->fr_icmp));
1836 i = 1;
1837 }
1838 }
1839 }
1840 }
1841 return i;
1842 }
1843
1844
1845 /* ------------------------------------------------------------------------ */
1846 /* Function: fr_scanlist */
1847 /* Returns: int - result flags of scanning filter list */
1848 /* Parameters: fin(I) - pointer to packet information */
1849 /* pass(I) - default result to return for filtering */
1850 /* */
1851 /* Check the input/output list of rules for a match to the current packet. */
1852 /* If a match is found, the value of fr_flags from the rule becomes the */
1853 /* return value and fin->fin_fr points to the matched rule. */
1854 /* */
1855 /* This function may be called recusively upto 16 times (limit inbuilt.) */
1856 /* When unwinding, it should finish up with fin_depth as 0. */
1857 /* */
1858 /* Could be per interface, but this gets real nasty when you don't have, */
1859 /* or can't easily change, the kernel source code to . */
1860 /* ------------------------------------------------------------------------ */
fr_scanlist(fin,pass)1861 int fr_scanlist(fin, pass)
1862 fr_info_t *fin;
1863 u_32_t pass;
1864 {
1865 int rulen, portcmp, off, logged, skip;
1866 struct frentry *fr, *fnext;
1867 u_32_t passt, passo;
1868 ipf_stack_t *ifs = fin->fin_ifs;
1869
1870 /*
1871 * Do not allow nesting deeper than 16 levels.
1872 */
1873 if (fin->fin_depth >= 16)
1874 return pass;
1875
1876 fr = fin->fin_fr;
1877
1878 /*
1879 * If there are no rules in this list, return now.
1880 */
1881 if (fr == NULL)
1882 return pass;
1883
1884 skip = 0;
1885 logged = 0;
1886 portcmp = 0;
1887 fin->fin_depth++;
1888 fin->fin_fr = NULL;
1889 off = fin->fin_off;
1890
1891 if ((fin->fin_flx & FI_TCPUDP) && (fin->fin_dlen > 3) && !off)
1892 portcmp = 1;
1893
1894 for (rulen = 0; fr; fr = fnext, rulen++) {
1895 fnext = fr->fr_next;
1896 if (skip != 0) {
1897 FR_VERBOSE(("%d (%#x)\n", skip, fr->fr_flags));
1898 skip--;
1899 continue;
1900 }
1901
1902 /*
1903 * In all checks below, a null (zero) value in the
1904 * filter struture is taken to mean a wildcard.
1905 *
1906 * check that we are working for the right interface
1907 */
1908 #ifdef _KERNEL
1909 if (fr->fr_ifa && fr->fr_ifa != fin->fin_ifp)
1910 continue;
1911 #else
1912 if (opts & (OPT_VERBOSE|OPT_DEBUG))
1913 printf("\n");
1914 FR_VERBOSE(("%c", FR_ISSKIP(pass) ? 's' :
1915 FR_ISPASS(pass) ? 'p' :
1916 FR_ISACCOUNT(pass) ? 'A' :
1917 FR_ISAUTH(pass) ? 'a' :
1918 (pass & FR_NOMATCH) ? 'n' :'b'));
1919 if (fr->fr_ifa && fr->fr_ifa != fin->fin_ifp)
1920 continue;
1921 FR_VERBOSE((":i"));
1922 #endif
1923
1924 switch (fr->fr_type)
1925 {
1926 case FR_T_IPF :
1927 case FR_T_IPF|FR_T_BUILTIN :
1928 if (fr_ipfcheck(fin, fr, portcmp))
1929 continue;
1930 break;
1931 #if defined(IPFILTER_BPF)
1932 case FR_T_BPFOPC :
1933 case FR_T_BPFOPC|FR_T_BUILTIN :
1934 {
1935 u_char *mc;
1936
1937 if (*fin->fin_mp == NULL)
1938 continue;
1939 if (fin->fin_v != fr->fr_v)
1940 continue;
1941 mc = (u_char *)fin->fin_m;
1942 if (!bpf_filter(fr->fr_data, mc, fin->fin_plen, 0))
1943 continue;
1944 break;
1945 }
1946 #endif
1947 case FR_T_CALLFUNC|FR_T_BUILTIN :
1948 {
1949 frentry_t *f;
1950
1951 f = (*fr->fr_func)(fin, &pass);
1952 if (f != NULL)
1953 fr = f;
1954 else
1955 continue;
1956 break;
1957 }
1958 default :
1959 break;
1960 }
1961
1962 if ((fin->fin_out == 0) && (fr->fr_nattag.ipt_num[0] != 0)) {
1963 if (fin->fin_nattag == NULL)
1964 continue;
1965 if (fr_matchtag(&fr->fr_nattag, fin->fin_nattag) == 0)
1966 continue;
1967 }
1968 FR_VERBOSE(("=%s.%d *", fr->fr_group, rulen));
1969
1970 passt = fr->fr_flags;
1971
1972 /*
1973 * Allowing a rule with the "keep state" flag set to match
1974 * packets that have been tagged "out of window" by the TCP
1975 * state tracking is foolish as the attempt to add a new
1976 * state entry to the table will fail.
1977 */
1978 if ((passt & FR_KEEPSTATE) && (fin->fin_flx & FI_OOW))
1979 continue;
1980
1981 /*
1982 * If the rule is a "call now" rule, then call the function
1983 * in the rule, if it exists and use the results from that.
1984 * If the function pointer is bad, just make like we ignore
1985 * it, except for increasing the hit counter.
1986 */
1987 IPF_BUMP(fr->fr_hits);
1988 fr->fr_bytes += (U_QUAD_T)fin->fin_plen;
1989 if ((passt & FR_CALLNOW) != 0) {
1990 if ((fr->fr_func != NULL) &&
1991 (fr->fr_func != (ipfunc_t)-1)) {
1992 frentry_t *frs;
1993
1994 frs = fin->fin_fr;
1995 fin->fin_fr = fr;
1996 fr = (*fr->fr_func)(fin, &passt);
1997 if (fr == NULL) {
1998 fin->fin_fr = frs;
1999 continue;
2000 }
2001 passt = fr->fr_flags;
2002 fin->fin_fr = fr;
2003 }
2004 } else {
2005 fin->fin_fr = fr;
2006 }
2007
2008 #ifdef IPFILTER_LOG
2009 /*
2010 * Just log this packet...
2011 */
2012 if ((passt & FR_LOGMASK) == FR_LOG) {
2013 if (ipflog(fin, passt) == -1) {
2014 if (passt & FR_LOGORBLOCK) {
2015 passt &= ~FR_CMDMASK;
2016 passt |= FR_BLOCK|FR_QUICK;
2017 }
2018 IPF_BUMP(ifs->ifs_frstats[fin->fin_out].fr_skip);
2019 }
2020 IPF_BUMP(ifs->ifs_frstats[fin->fin_out].fr_pkl);
2021 logged = 1;
2022 }
2023 #endif /* IPFILTER_LOG */
2024 passo = pass;
2025 if (FR_ISSKIP(passt))
2026 skip = fr->fr_arg;
2027 else if ((passt & FR_LOGMASK) != FR_LOG)
2028 pass = passt;
2029 if (passt & (FR_RETICMP|FR_FAKEICMP))
2030 fin->fin_icode = fr->fr_icode;
2031 FR_DEBUG(("pass %#x\n", pass));
2032 fin->fin_rule = rulen;
2033 (void) strncpy(fin->fin_group, fr->fr_group, FR_GROUPLEN);
2034 if (fr->fr_grp != NULL) {
2035 fin->fin_fr = *fr->fr_grp;
2036 pass = fr_scanlist(fin, pass);
2037 if (fin->fin_fr == NULL) {
2038 fin->fin_rule = rulen;
2039 (void) strncpy(fin->fin_group, fr->fr_group,
2040 FR_GROUPLEN);
2041 fin->fin_fr = fr;
2042 }
2043 if (fin->fin_flx & FI_DONTCACHE)
2044 logged = 1;
2045 }
2046
2047 if (pass & FR_QUICK) {
2048 /*
2049 * Finally, if we've asked to track state for this
2050 * packet, set it up. Add state for "quick" rules
2051 * here so that if the action fails we can consider
2052 * the rule to "not match" and keep on processing
2053 * filter rules.
2054 */
2055 if ((pass & FR_KEEPSTATE) &&
2056 !(fin->fin_flx & FI_STATE)) {
2057 int out = fin->fin_out;
2058
2059 if (fr_addstate(fin, NULL, 0) != NULL) {
2060 IPF_BUMP(ifs->ifs_frstats[out].fr_ads);
2061 } else {
2062 IPF_BUMP(ifs->ifs_frstats[out].fr_bads);
2063 pass = passo;
2064 continue;
2065 }
2066 }
2067 break;
2068 }
2069 }
2070 if (logged)
2071 fin->fin_flx |= FI_DONTCACHE;
2072 fin->fin_depth--;
2073 return pass;
2074 }
2075
2076
2077 /* ------------------------------------------------------------------------ */
2078 /* Function: fr_acctpkt */
2079 /* Returns: frentry_t* - always returns NULL */
2080 /* Parameters: fin(I) - pointer to packet information */
2081 /* passp(IO) - pointer to current/new filter decision (unused) */
2082 /* */
2083 /* Checks a packet against accounting rules, if there are any for the given */
2084 /* IP protocol version. */
2085 /* */
2086 /* N.B.: this function returns NULL to match the prototype used by other */
2087 /* functions called from the IPFilter "mainline" in fr_check(). */
2088 /* ------------------------------------------------------------------------ */
fr_acctpkt(fin,passp)2089 frentry_t *fr_acctpkt(fin, passp)
2090 fr_info_t *fin;
2091 u_32_t *passp;
2092 {
2093 char group[FR_GROUPLEN];
2094 frentry_t *fr, *frsave;
2095 u_32_t pass, rulen;
2096 ipf_stack_t *ifs = fin->fin_ifs;
2097
2098 passp = passp;
2099 #ifdef USE_INET6
2100 if (fin->fin_v == 6)
2101 fr = ifs->ifs_ipacct6[fin->fin_out][ifs->ifs_fr_active];
2102 else
2103 #endif
2104 fr = ifs->ifs_ipacct[fin->fin_out][ifs->ifs_fr_active];
2105
2106 if (fr != NULL) {
2107 frsave = fin->fin_fr;
2108 bcopy(fin->fin_group, group, FR_GROUPLEN);
2109 rulen = fin->fin_rule;
2110 fin->fin_fr = fr;
2111 pass = fr_scanlist(fin, FR_NOMATCH);
2112 if (FR_ISACCOUNT(pass)) {
2113 IPF_BUMP(ifs->ifs_frstats[0].fr_acct);
2114 }
2115 fin->fin_fr = frsave;
2116 bcopy(group, fin->fin_group, FR_GROUPLEN);
2117 fin->fin_rule = rulen;
2118 }
2119 return NULL;
2120 }
2121
2122
2123 /* ------------------------------------------------------------------------ */
2124 /* Function: fr_firewall */
2125 /* Returns: frentry_t* - returns pointer to matched rule, if no matches */
2126 /* were found, returns NULL. */
2127 /* Parameters: fin(I) - pointer to packet information */
2128 /* passp(IO) - pointer to current/new filter decision (unused) */
2129 /* */
2130 /* Applies an appropriate set of firewall rules to the packet, to see if */
2131 /* there are any matches. The first check is to see if a match can be seen */
2132 /* in the cache. If not, then search an appropriate list of rules. Once a */
2133 /* matching rule is found, take any appropriate actions as defined by the */
2134 /* rule - except logging. */
2135 /* ------------------------------------------------------------------------ */
fr_firewall(fin,passp)2136 static frentry_t *fr_firewall(fin, passp)
2137 fr_info_t *fin;
2138 u_32_t *passp;
2139 {
2140 frentry_t *fr;
2141 fr_info_t *fc;
2142 u_32_t pass;
2143 int out;
2144 ipf_stack_t *ifs = fin->fin_ifs;
2145
2146 out = fin->fin_out;
2147 pass = *passp;
2148
2149 #ifdef USE_INET6
2150 if (fin->fin_v == 6)
2151 fin->fin_fr = ifs->ifs_ipfilter6[out][ifs->ifs_fr_active];
2152 else
2153 #endif
2154 fin->fin_fr = ifs->ifs_ipfilter[out][ifs->ifs_fr_active];
2155
2156 /*
2157 * If there are no rules loaded skip all checks and return.
2158 */
2159 if (fin->fin_fr == NULL) {
2160
2161 if ((pass & FR_NOMATCH)) {
2162 IPF_BUMP(ifs->ifs_frstats[out].fr_nom);
2163 }
2164
2165 return (NULL);
2166 }
2167
2168 fc = &ifs->ifs_frcache[out][CACHE_HASH(fin)];
2169 READ_ENTER(&ifs->ifs_ipf_frcache);
2170 if (!bcmp((char *)fin, (char *)fc, FI_CSIZE)) {
2171 /*
2172 * copy cached data so we can unlock the mutexes earlier.
2173 */
2174 bcopy((char *)fc, (char *)fin, FI_COPYSIZE);
2175 RWLOCK_EXIT(&ifs->ifs_ipf_frcache);
2176 IPF_BUMP(ifs->ifs_frstats[out].fr_chit);
2177
2178 if ((fr = fin->fin_fr) != NULL) {
2179 IPF_BUMP(fr->fr_hits);
2180 fr->fr_bytes += (U_QUAD_T)fin->fin_plen;
2181 pass = fr->fr_flags;
2182 }
2183 } else {
2184 RWLOCK_EXIT(&ifs->ifs_ipf_frcache);
2185
2186 pass = fr_scanlist(fin, ifs->ifs_fr_pass);
2187
2188 if (((pass & FR_KEEPSTATE) == 0) &&
2189 ((fin->fin_flx & FI_DONTCACHE) == 0)) {
2190 WRITE_ENTER(&ifs->ifs_ipf_frcache);
2191 bcopy((char *)fin, (char *)fc, FI_COPYSIZE);
2192 RWLOCK_EXIT(&ifs->ifs_ipf_frcache);
2193 }
2194
2195 fr = fin->fin_fr;
2196 }
2197
2198 if ((pass & FR_NOMATCH)) {
2199 IPF_BUMP(ifs->ifs_frstats[out].fr_nom);
2200 }
2201
2202 /*
2203 * Apply packets per second rate-limiting to a rule as required.
2204 */
2205 if ((fr != NULL) && (fr->fr_pps != 0) &&
2206 !ppsratecheck(&fr->fr_lastpkt, &fr->fr_curpps, fr->fr_pps)) {
2207 pass &= ~(FR_CMDMASK|FR_DUP|FR_RETICMP|FR_RETRST);
2208 pass |= FR_BLOCK;
2209 IPF_BUMP(ifs->ifs_frstats[out].fr_ppshit);
2210 }
2211
2212 /*
2213 * If we fail to add a packet to the authorization queue, then we
2214 * drop the packet later. However, if it was added then pretend
2215 * we've dropped it already.
2216 */
2217 if (FR_ISAUTH(pass)) {
2218 if (fr_newauth(fin->fin_m, fin) != 0) {
2219 #ifdef _KERNEL
2220 fin->fin_m = *fin->fin_mp = NULL;
2221 #else
2222 ;
2223 #endif
2224 fin->fin_error = 0;
2225 } else
2226 fin->fin_error = ENOSPC;
2227 }
2228
2229 if ((fr != NULL) && (fr->fr_func != NULL) &&
2230 (fr->fr_func != (ipfunc_t)-1) && !(pass & FR_CALLNOW))
2231 (void) (*fr->fr_func)(fin, &pass);
2232
2233 /*
2234 * If a rule is a pre-auth rule, check again in the list of rules
2235 * loaded for authenticated use. It does not particulary matter
2236 * if this search fails because a "preauth" result, from a rule,
2237 * is treated as "not a pass", hence the packet is blocked.
2238 */
2239 if (FR_ISPREAUTH(pass)) {
2240 if ((fin->fin_fr = ifs->ifs_ipauth) != NULL)
2241 pass = fr_scanlist(fin, ifs->ifs_fr_pass);
2242 }
2243
2244 /*
2245 * If the rule has "keep frag" and the packet is actually a fragment,
2246 * then create a fragment state entry.
2247 */
2248 if ((pass & (FR_KEEPFRAG|FR_KEEPSTATE)) == FR_KEEPFRAG) {
2249 if (fin->fin_flx & FI_FRAG) {
2250 if (fr_newfrag(fin, pass) == -1) {
2251 IPF_BUMP(ifs->ifs_frstats[out].fr_bnfr);
2252 } else {
2253 IPF_BUMP(ifs->ifs_frstats[out].fr_nfr);
2254 }
2255 } else {
2256 IPF_BUMP(ifs->ifs_frstats[out].fr_cfr);
2257 }
2258 }
2259
2260 /*
2261 * Finally, if we've asked to track state for this packet, set it up.
2262 */
2263 if ((pass & FR_KEEPSTATE) && !(fin->fin_flx & FI_STATE)) {
2264 if (fr_addstate(fin, NULL, 0) != NULL) {
2265 IPF_BUMP(ifs->ifs_frstats[out].fr_ads);
2266 } else {
2267 IPF_BUMP(ifs->ifs_frstats[out].fr_bads);
2268 if (FR_ISPASS(pass)) {
2269 pass &= ~FR_CMDMASK;
2270 pass |= FR_BLOCK;
2271 }
2272 }
2273 }
2274
2275 fr = fin->fin_fr;
2276
2277 if (passp != NULL)
2278 *passp = pass;
2279
2280 return fr;
2281 }
2282
2283 /* ------------------------------------------------------------------------ */
2284 /* Function: fr_check */
2285 /* Returns: int - 0 == packet allowed through, */
2286 /* User space: */
2287 /* -1 == packet blocked */
2288 /* 1 == packet not matched */
2289 /* -2 == requires authentication */
2290 /* Kernel: */
2291 /* > 0 == filter error # for packet */
2292 /* Parameters: ip(I) - pointer to start of IPv4/6 packet */
2293 /* hlen(I) - length of header */
2294 /* ifp(I) - pointer to interface this packet is on */
2295 /* out(I) - 0 == packet going in, 1 == packet going out */
2296 /* mp(IO) - pointer to caller's buffer pointer that holds this */
2297 /* IP packet. */
2298 /* Solaris & HP-UX ONLY : */
2299 /* qpi(I) - pointer to STREAMS queue information for this */
2300 /* interface & direction. */
2301 /* */
2302 /* fr_check() is the master function for all IPFilter packet processing. */
2303 /* It orchestrates: Network Address Translation (NAT), checking for packet */
2304 /* authorisation (or pre-authorisation), presence of related state info., */
2305 /* generating log entries, IP packet accounting, routing of packets as */
2306 /* directed by firewall rules and of course whether or not to allow the */
2307 /* packet to be further processed by the kernel. */
2308 /* */
2309 /* For packets blocked, the contents of "mp" will be NULL'd and the buffer */
2310 /* freed. Packets passed may be returned with the pointer pointed to by */
2311 /* by "mp" changed to a new buffer. */
2312 /* ------------------------------------------------------------------------ */
fr_check(ip,hlen,ifp,out,qif,mp,ifs)2313 int fr_check(ip, hlen, ifp, out
2314 #if defined(_KERNEL) && defined(MENTAT)
2315 , qif, mp, ifs)
2316 void *qif;
2317 #else
2318 , mp, ifs)
2319 #endif
2320 mb_t **mp;
2321 ip_t *ip;
2322 int hlen;
2323 void *ifp;
2324 int out;
2325 ipf_stack_t *ifs;
2326 {
2327 /*
2328 * The above really sucks, but short of writing a diff
2329 */
2330 fr_info_t frinfo;
2331 fr_info_t *fin = &frinfo;
2332 u_32_t pass;
2333 frentry_t *fr = NULL;
2334 int v = IP_V(ip);
2335 mb_t *mc = NULL;
2336 mb_t *m;
2337 #ifdef USE_INET6
2338 ip6_t *ip6;
2339 #endif
2340 #ifdef _KERNEL
2341 # ifdef MENTAT
2342 qpktinfo_t *qpi = qif;
2343 #endif
2344 #endif
2345
2346 SPL_INT(s);
2347 pass = ifs->ifs_fr_pass;
2348
2349 /*
2350 * The first part of fr_check() deals with making sure that what goes
2351 * into the filtering engine makes some sense. Information about the
2352 * the packet is distilled, collected into a fr_info_t structure and
2353 * the an attempt to ensure the buffer the packet is in is big enough
2354 * to hold all the required packet headers.
2355 */
2356 #ifdef _KERNEL
2357 # ifdef MENTAT
2358 if (!OK_32PTR(ip))
2359 return 2;
2360 # endif
2361
2362
2363 if (ifs->ifs_fr_running <= 0) {
2364 return 0;
2365 }
2366
2367 bzero((char *)fin, sizeof(*fin));
2368
2369 # ifdef MENTAT
2370 fin->fin_flx = qpi->qpi_flags & (FI_NOCKSUM|FI_MBCAST|FI_MULTICAST|
2371 FI_BROADCAST);
2372 m = qpi->qpi_m;
2373 fin->fin_qfm = m;
2374 fin->fin_qpi = qpi;
2375 # else /* MENTAT */
2376
2377 m = *mp;
2378
2379 # if defined(M_MCAST)
2380 if ((m->m_flags & M_MCAST) != 0)
2381 fin->fin_flx |= FI_MBCAST|FI_MULTICAST;
2382 # endif
2383 # if defined(M_MLOOP)
2384 if ((m->m_flags & M_MLOOP) != 0)
2385 fin->fin_flx |= FI_MBCAST|FI_MULTICAST;
2386 # endif
2387 # if defined(M_BCAST)
2388 if ((m->m_flags & M_BCAST) != 0)
2389 fin->fin_flx |= FI_MBCAST|FI_BROADCAST;
2390 # endif
2391 # ifdef M_CANFASTFWD
2392 /*
2393 * XXX For now, IP Filter and fast-forwarding of cached flows
2394 * XXX are mutually exclusive. Eventually, IP Filter should
2395 * XXX get a "can-fast-forward" filter rule.
2396 */
2397 m->m_flags &= ~M_CANFASTFWD;
2398 # endif /* M_CANFASTFWD */
2399 # ifdef CSUM_DELAY_DATA
2400 /*
2401 * disable delayed checksums.
2402 */
2403 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
2404 in_delayed_cksum(m);
2405 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
2406 }
2407 # endif /* CSUM_DELAY_DATA */
2408 # endif /* MENTAT */
2409 #else
2410
2411 bzero((char *)fin, sizeof(*fin));
2412 m = *mp;
2413 #endif /* _KERNEL */
2414
2415 fin->fin_v = v;
2416 fin->fin_m = m;
2417 fin->fin_ip = ip;
2418 fin->fin_mp = mp;
2419 fin->fin_out = out;
2420 fin->fin_ifp = ifp;
2421 fin->fin_error = ENETUNREACH;
2422 fin->fin_hlen = (u_short)hlen;
2423 fin->fin_dp = (char *)ip + hlen;
2424 fin->fin_ipoff = (char *)ip - MTOD(m, char *);
2425 fin->fin_ifs = ifs;
2426
2427 SPL_NET(s);
2428
2429 #ifdef USE_INET6
2430 if (v == 6) {
2431 IPF_BUMP(ifs->ifs_frstats[out].fr_ipv6);
2432 /*
2433 * Jumbo grams are quite likely too big for internal buffer
2434 * structures to handle comfortably, for now, so just drop
2435 * them.
2436 */
2437 ip6 = (ip6_t *)ip;
2438 fin->fin_plen = ntohs(ip6->ip6_plen);
2439 if (fin->fin_plen == 0) {
2440 READ_ENTER(&ifs->ifs_ipf_mutex);
2441 pass = FR_BLOCK|FR_NOMATCH;
2442 goto filtered;
2443 }
2444 fin->fin_plen += sizeof(ip6_t);
2445 } else
2446 #endif
2447 {
2448 #if (OpenBSD >= 200311) && defined(_KERNEL)
2449 ip->ip_len = ntohs(ip->ip_len);
2450 ip->ip_off = ntohs(ip->ip_off);
2451 #endif
2452 fin->fin_plen = ip->ip_len;
2453 }
2454
2455 if (fr_makefrip(hlen, ip, fin) == -1) {
2456 READ_ENTER(&ifs->ifs_ipf_mutex);
2457 pass = FR_BLOCK;
2458 goto filtered;
2459 }
2460
2461 /*
2462 * For at least IPv6 packets, if a m_pullup() fails then this pointer
2463 * becomes NULL and so we have no packet to free.
2464 */
2465 if (*fin->fin_mp == NULL)
2466 goto finished;
2467
2468 if (!out) {
2469 if (v == 4) {
2470 #ifdef _KERNEL
2471 if (ifs->ifs_fr_chksrc && !fr_verifysrc(fin)) {
2472 IPF_BUMP(ifs->ifs_frstats[0].fr_badsrc);
2473 fin->fin_flx |= FI_BADSRC;
2474 }
2475 #endif
2476 if (fin->fin_ip->ip_ttl < ifs->ifs_fr_minttl) {
2477 IPF_BUMP(ifs->ifs_frstats[0].fr_badttl);
2478 fin->fin_flx |= FI_LOWTTL;
2479 }
2480 }
2481 #ifdef USE_INET6
2482 else if (v == 6) {
2483 ip6 = (ip6_t *)ip;
2484 #ifdef _KERNEL
2485 if (ifs->ifs_fr_chksrc && !fr_verifysrc(fin)) {
2486 IPF_BUMP(ifs->ifs_frstats[0].fr_badsrc);
2487 fin->fin_flx |= FI_BADSRC;
2488 }
2489 #endif
2490 if (ip6->ip6_hlim < ifs->ifs_fr_minttl) {
2491 IPF_BUMP(ifs->ifs_frstats[0].fr_badttl);
2492 fin->fin_flx |= FI_LOWTTL;
2493 }
2494 }
2495 #endif
2496 }
2497
2498 if (fin->fin_flx & FI_SHORT) {
2499 IPF_BUMP(ifs->ifs_frstats[out].fr_short);
2500 }
2501
2502 READ_ENTER(&ifs->ifs_ipf_mutex);
2503
2504 /*
2505 * Check auth now. This, combined with the check below to see if apass
2506 * is 0 is to ensure that we don't count the packet twice, which can
2507 * otherwise occur when we reprocess it. As it is, we only count it
2508 * after it has no auth. table matchup. This also stops NAT from
2509 * occuring until after the packet has been auth'd.
2510 */
2511 fr = fr_checkauth(fin, &pass);
2512 if (!out) {
2513 switch (fin->fin_v)
2514 {
2515 case 4 :
2516 if (fr_checknatin(fin, &pass) == -1) {
2517 RWLOCK_EXIT(&ifs->ifs_ipf_mutex);
2518 goto finished;
2519 }
2520 break;
2521 #ifdef USE_INET6
2522 case 6 :
2523 if (fr_checknat6in(fin, &pass) == -1) {
2524 RWLOCK_EXIT(&ifs->ifs_ipf_mutex);
2525 goto finished;
2526 }
2527 break;
2528 #endif
2529 default :
2530 break;
2531 }
2532 }
2533 if (!out)
2534 (void) fr_acctpkt(fin, NULL);
2535
2536 if (fr == NULL)
2537 if ((fin->fin_flx & (FI_FRAG|FI_BAD)) == FI_FRAG)
2538 fr = fr_knownfrag(fin, &pass);
2539 if (fr == NULL)
2540 fr = fr_checkstate(fin, &pass);
2541
2542 if ((pass & FR_NOMATCH) || (fr == NULL))
2543 fr = fr_firewall(fin, &pass);
2544
2545 fin->fin_fr = fr;
2546
2547 /*
2548 * Only count/translate packets which will be passed on, out the
2549 * interface.
2550 */
2551 if (out && FR_ISPASS(pass)) {
2552 (void) fr_acctpkt(fin, NULL);
2553
2554 switch (fin->fin_v)
2555 {
2556 case 4 :
2557 if (fr_checknatout(fin, &pass) == -1) {
2558 RWLOCK_EXIT(&ifs->ifs_ipf_mutex);
2559 goto finished;
2560 }
2561 break;
2562 #ifdef USE_INET6
2563 case 6 :
2564 if (fr_checknat6out(fin, &pass) == -1) {
2565 RWLOCK_EXIT(&ifs->ifs_ipf_mutex);
2566 goto finished;
2567 }
2568 break;
2569 #endif
2570 default :
2571 break;
2572 }
2573
2574 if ((ifs->ifs_fr_update_ipid != 0) && (v == 4)) {
2575 if (fr_updateipid(fin) == -1) {
2576 IPF_BUMP(ifs->ifs_frstats[1].fr_ipud);
2577 pass &= ~FR_CMDMASK;
2578 pass |= FR_BLOCK;
2579 } else {
2580 IPF_BUMP(ifs->ifs_frstats[0].fr_ipud);
2581 }
2582 }
2583 }
2584
2585 #ifdef IPFILTER_LOG
2586 if ((ifs->ifs_fr_flags & FF_LOGGING) || (pass & FR_LOGMASK)) {
2587 (void) fr_dolog(fin, &pass);
2588 }
2589 #endif
2590
2591 /*
2592 * The FI_STATE flag is cleared here so that calling fr_checkstate
2593 * will work when called from inside of fr_fastroute. Although
2594 * there is a similar flag, FI_NATED, for NAT, it does have the same
2595 * impact on code execution.
2596 */
2597 fin->fin_flx &= ~FI_STATE;
2598
2599 /*
2600 * Only allow FR_DUP to work if a rule matched - it makes no sense to
2601 * set FR_DUP as a "default" as there are no instructions about where
2602 * to send the packet. Use fin_m here because it may have changed
2603 * (without an update of 'm') in prior processing.
2604 */
2605 if ((fr != NULL) && (pass & FR_DUP)) {
2606 mc = M_DUPLICATE(fin->fin_m);
2607 #ifdef _KERNEL
2608 mc->b_rptr += fin->fin_ipoff;
2609 #endif
2610 }
2611
2612 /*
2613 * We don't want to send RST for packets, which are going to be
2614 * dropped, just because they don't fit into TCP window. Those packets
2615 * will be dropped silently. In other words, we want to drop packet,
2616 * while keeping session alive.
2617 */
2618 if ((pass & (FR_RETRST|FR_RETICMP)) && ((fin->fin_flx & FI_OOW) == 0)) {
2619 /*
2620 * Should we return an ICMP packet to indicate error
2621 * status passing through the packet filter ?
2622 * WARNING: ICMP error packets AND TCP RST packets should
2623 * ONLY be sent in repsonse to incoming packets. Sending them
2624 * in response to outbound packets can result in a panic on
2625 * some operating systems.
2626 */
2627 if (!out) {
2628 if (pass & FR_RETICMP) {
2629 int dst;
2630
2631 if ((pass & FR_RETMASK) == FR_FAKEICMP)
2632 dst = 1;
2633 else
2634 dst = 0;
2635 #if defined(_KERNEL) && (SOLARIS2 >= 10)
2636 /*
2637 * Assume it's possible to enter insane rule:
2638 * pass return-icmp in proto udp ...
2639 * then we have no other option than to forward
2640 * packet on loopback and give up any attempt
2641 * to create a fake response.
2642 */
2643 if (IPF_IS_LOOPBACK(qpi->qpi_flags) &&
2644 FR_ISBLOCK(pass)) {
2645
2646 if (fr_make_icmp(fin) == 0) {
2647 IPF_BUMP(
2648 ifs->ifs_frstats[out].fr_ret);
2649 }
2650 /*
2651 * we drop packet silently in case we
2652 * failed assemble fake response for it
2653 */
2654 else if (*mp != NULL) {
2655 FREE_MB_T(*mp);
2656 m = *mp = NULL;
2657 }
2658
2659 IPF_BUMP(
2660 ifs->ifs_frstats[out].fr_block);
2661 RWLOCK_EXIT(&ifs->ifs_ipf_mutex);
2662
2663 return (0);
2664 }
2665 #endif /* _KERNEL && SOLARIS2 >= 10 */
2666
2667 (void) fr_send_icmp_err(ICMP_UNREACH, fin, dst);
2668 IPF_BUMP(ifs->ifs_frstats[out].fr_ret);
2669
2670 } else if (((pass & FR_RETMASK) == FR_RETRST) &&
2671 !(fin->fin_flx & FI_SHORT)) {
2672
2673 #if defined(_KERNEL) && (SOLARIS2 >= 10)
2674 /*
2675 * Assume it's possible to enter insane rule:
2676 * pass return-rst in proto tcp ...
2677 * then we have no other option than to forward
2678 * packet on loopback and give up any attempt
2679 * to create a fake response.
2680 */
2681 if (IPF_IS_LOOPBACK(qpi->qpi_flags) &&
2682 FR_ISBLOCK(pass)) {
2683 if (fr_make_rst(fin) == 0) {
2684 IPF_BUMP(
2685 ifs->ifs_frstats[out].fr_ret);
2686 }
2687 else if (mp != NULL) {
2688 /*
2689 * we drop packet silently in case we
2690 * failed assemble fake response for it
2691 */
2692 FREE_MB_T(*mp);
2693 m = *mp = NULL;
2694 }
2695
2696 IPF_BUMP(
2697 ifs->ifs_frstats[out].fr_block);
2698 RWLOCK_EXIT(&ifs->ifs_ipf_mutex);
2699
2700 return (0);
2701 }
2702 #endif /* _KERNEL && _SOLARIS2 >= 10 */
2703 if (fr_send_reset(fin) == 0) {
2704 IPF_BUMP(ifs->ifs_frstats[1].fr_ret);
2705 }
2706 }
2707 } else {
2708 if (pass & FR_RETRST)
2709 fin->fin_error = ECONNRESET;
2710 }
2711 }
2712
2713 /*
2714 * If we didn't drop off the bottom of the list of rules (and thus
2715 * the 'current' rule fr is not NULL), then we may have some extra
2716 * instructions about what to do with a packet.
2717 * Once we're finished return to our caller, freeing the packet if
2718 * we are dropping it (* BSD ONLY *).
2719 * Reassign m from fin_m as we may have a new buffer, now.
2720 */
2721 filtered:
2722 m = fin->fin_m;
2723
2724 if (fr != NULL) {
2725 frdest_t *fdp;
2726
2727 fdp = &fr->fr_tifs[fin->fin_rev];
2728
2729 if (!out && (pass & FR_FASTROUTE)) {
2730 /*
2731 * For fastroute rule, no destioation interface defined
2732 * so pass NULL as the frdest_t parameter
2733 */
2734 (void) fr_fastroute(m, mp, fin, NULL);
2735 m = *mp = NULL;
2736 } else if ((fdp->fd_ifp != NULL) &&
2737 (fdp->fd_ifp != (struct ifnet *)-1)) {
2738 /* this is for to rules: */
2739 (void) fr_fastroute(m, mp, fin, fdp);
2740 m = *mp = NULL;
2741 }
2742
2743 /*
2744 * Send a duplicated packet.
2745 */
2746 if (mc != NULL) {
2747 #if defined(_KERNEL) && (SOLARIS2 >= 10)
2748 /*
2749 * We are going to compute chksum for copies of loopback packets
2750 * only. IP stack does not compute chksums at all for loopback
2751 * packets. We want to get it fixed in their copies, since those
2752 * are going to be sent to network.
2753 */
2754 if (IPF_IS_LOOPBACK(qpi->qpi_flags))
2755 fr_calc_chksum(fin, mc);
2756 #endif
2757 (void) fr_fastroute(mc, &mc, fin, &fr->fr_dif);
2758 }
2759 }
2760
2761 if (FR_ISBLOCK(pass) && (fin->fin_flx & FI_NEWNAT))
2762 nat_uncreate(fin);
2763
2764 /*
2765 * This late because the likes of fr_fastroute() use fin_fr.
2766 */
2767 RWLOCK_EXIT(&ifs->ifs_ipf_mutex);
2768
2769 finished:
2770 if (!FR_ISPASS(pass)) {
2771 IPF_BUMP(ifs->ifs_frstats[out].fr_block);
2772 if (*mp != NULL) {
2773 FREE_MB_T(*mp);
2774 m = *mp = NULL;
2775 }
2776 } else {
2777 IPF_BUMP(ifs->ifs_frstats[out].fr_pass);
2778 #if defined(_KERNEL) && defined(__sgi)
2779 if ((fin->fin_hbuf != NULL) &&
2780 (mtod(fin->fin_m, struct ip *) != fin->fin_ip)) {
2781 COPYBACK(m, 0, fin->fin_plen, fin->fin_hbuf);
2782 }
2783 #endif
2784 }
2785
2786 SPL_X(s);
2787
2788 #ifdef _KERNEL
2789 # if OpenBSD >= 200311
2790 if (FR_ISPASS(pass) && (v == 4)) {
2791 ip = fin->fin_ip;
2792 ip->ip_len = ntohs(ip->ip_len);
2793 ip->ip_off = ntohs(ip->ip_off);
2794 }
2795 # endif
2796 return (FR_ISPASS(pass)) ? 0 : fin->fin_error;
2797 #else /* _KERNEL */
2798 FR_VERBOSE(("fin_flx %#x pass %#x ", fin->fin_flx, pass));
2799 if ((pass & FR_NOMATCH) != 0)
2800 return 1;
2801
2802 if ((pass & FR_RETMASK) != 0)
2803 switch (pass & FR_RETMASK)
2804 {
2805 case FR_RETRST :
2806 return 3;
2807 case FR_RETICMP :
2808 return 4;
2809 case FR_FAKEICMP :
2810 return 5;
2811 }
2812
2813 switch (pass & FR_CMDMASK)
2814 {
2815 case FR_PASS :
2816 return 0;
2817 case FR_BLOCK :
2818 return -1;
2819 case FR_AUTH :
2820 return -2;
2821 case FR_ACCOUNT :
2822 return -3;
2823 case FR_PREAUTH :
2824 return -4;
2825 }
2826 return 2;
2827 #endif /* _KERNEL */
2828 }
2829
2830
2831 #ifdef IPFILTER_LOG
2832 /* ------------------------------------------------------------------------ */
2833 /* Function: fr_dolog */
2834 /* Returns: frentry_t* - returns contents of fin_fr (no change made) */
2835 /* Parameters: fin(I) - pointer to packet information */
2836 /* passp(IO) - pointer to current/new filter decision (unused) */
2837 /* */
2838 /* Checks flags set to see how a packet should be logged, if it is to be */
2839 /* logged. Adjust statistics based on its success or not. */
2840 /* ------------------------------------------------------------------------ */
fr_dolog(fin,passp)2841 frentry_t *fr_dolog(fin, passp)
2842 fr_info_t *fin;
2843 u_32_t *passp;
2844 {
2845 u_32_t pass;
2846 int out;
2847 ipf_stack_t *ifs = fin->fin_ifs;
2848
2849 out = fin->fin_out;
2850 pass = *passp;
2851
2852 if ((ifs->ifs_fr_flags & FF_LOGNOMATCH) && (pass & FR_NOMATCH)) {
2853 pass |= FF_LOGNOMATCH;
2854 IPF_BUMP(ifs->ifs_frstats[out].fr_npkl);
2855 goto logit;
2856 } else if (((pass & FR_LOGMASK) == FR_LOGP) ||
2857 (FR_ISPASS(pass) && (ifs->ifs_fr_flags & FF_LOGPASS))) {
2858 if ((pass & FR_LOGMASK) != FR_LOGP)
2859 pass |= FF_LOGPASS;
2860 IPF_BUMP(ifs->ifs_frstats[out].fr_ppkl);
2861 goto logit;
2862 } else if (((pass & FR_LOGMASK) == FR_LOGB) ||
2863 (FR_ISBLOCK(pass) && (ifs->ifs_fr_flags & FF_LOGBLOCK))) {
2864 if ((pass & FR_LOGMASK) != FR_LOGB)
2865 pass |= FF_LOGBLOCK;
2866 IPF_BUMP(ifs->ifs_frstats[out].fr_bpkl);
2867 logit:
2868 if (ipflog(fin, pass) == -1) {
2869 IPF_BUMP(ifs->ifs_frstats[out].fr_skip);
2870
2871 /*
2872 * If the "or-block" option has been used then
2873 * block the packet if we failed to log it.
2874 */
2875 if ((pass & FR_LOGORBLOCK) &&
2876 FR_ISPASS(pass)) {
2877 pass &= ~FR_CMDMASK;
2878 pass |= FR_BLOCK;
2879 }
2880 }
2881 *passp = pass;
2882 }
2883
2884 return fin->fin_fr;
2885 }
2886 #endif /* IPFILTER_LOG */
2887
2888
2889 /* ------------------------------------------------------------------------ */
2890 /* Function: ipf_cksum */
2891 /* Returns: u_short - IP header checksum */
2892 /* Parameters: addr(I) - pointer to start of buffer to checksum */
2893 /* len(I) - length of buffer in bytes */
2894 /* */
2895 /* Calculate the two's complement 16 bit checksum of the buffer passed. */
2896 /* */
2897 /* N.B.: addr should be 16bit aligned. */
2898 /* ------------------------------------------------------------------------ */
ipf_cksum(addr,len)2899 u_short ipf_cksum(addr, len)
2900 u_short *addr;
2901 int len;
2902 {
2903 u_32_t sum = 0;
2904
2905 for (sum = 0; len > 1; len -= 2)
2906 sum += *addr++;
2907
2908 /* mop up an odd byte, if necessary */
2909 if (len == 1)
2910 sum += *(u_char *)addr;
2911
2912 /*
2913 * add back carry outs from top 16 bits to low 16 bits
2914 */
2915 sum = (sum >> 16) + (sum & 0xffff); /* add hi 16 to low 16 */
2916 sum += (sum >> 16); /* add carry */
2917 return (u_short)(~sum);
2918 }
2919
2920
2921 /* ------------------------------------------------------------------------ */
2922 /* Function: fr_cksum */
2923 /* Returns: u_short - layer 4 checksum */
2924 /* Parameters: m(I ) - pointer to buffer holding packet */
2925 /* ip(I) - pointer to IP header */
2926 /* l4proto(I) - protocol to caclulate checksum for */
2927 /* l4hdr(I) - pointer to layer 4 header */
2928 /* */
2929 /* Calculates the TCP checksum for the packet held in "m", using the data */
2930 /* in the IP header "ip" to seed it. */
2931 /* */
2932 /* NB: This function assumes we've pullup'd enough for all of the IP header */
2933 /* and the TCP header. We also assume that data blocks aren't allocated in */
2934 /* odd sizes. */
2935 /* */
2936 /* Expects ip_len to be in host byte order when called. */
2937 /* ------------------------------------------------------------------------ */
fr_cksum(m,ip,l4proto,l4hdr)2938 u_short fr_cksum(m, ip, l4proto, l4hdr)
2939 mb_t *m;
2940 ip_t *ip;
2941 int l4proto;
2942 void *l4hdr;
2943 {
2944 u_short *sp, slen, sumsave, l4hlen, *csump;
2945 u_int sum, sum2;
2946 int hlen;
2947 #ifdef USE_INET6
2948 ip6_t *ip6;
2949 #endif
2950
2951 csump = NULL;
2952 sumsave = 0;
2953 l4hlen = 0;
2954 sp = NULL;
2955 slen = 0;
2956 hlen = 0;
2957 sum = 0;
2958
2959 /*
2960 * Add up IP Header portion
2961 */
2962 #ifdef USE_INET6
2963 if (IP_V(ip) == 4) {
2964 #endif
2965 hlen = IP_HL(ip) << 2;
2966 slen = ip->ip_len - hlen;
2967 sum = htons((u_short)l4proto);
2968 sum += htons(slen);
2969 sp = (u_short *)&ip->ip_src;
2970 sum += *sp++; /* ip_src */
2971 sum += *sp++;
2972 sum += *sp++; /* ip_dst */
2973 sum += *sp++;
2974 #ifdef USE_INET6
2975 } else if (IP_V(ip) == 6) {
2976 ip6 = (ip6_t *)ip;
2977 hlen = sizeof(*ip6);
2978 slen = ntohs(ip6->ip6_plen);
2979 sum = htons((u_short)l4proto);
2980 sum += htons(slen);
2981 sp = (u_short *)&ip6->ip6_src;
2982 sum += *sp++; /* ip6_src */
2983 sum += *sp++;
2984 sum += *sp++;
2985 sum += *sp++;
2986 sum += *sp++;
2987 sum += *sp++;
2988 sum += *sp++;
2989 sum += *sp++;
2990 sum += *sp++; /* ip6_dst */
2991 sum += *sp++;
2992 sum += *sp++;
2993 sum += *sp++;
2994 sum += *sp++;
2995 sum += *sp++;
2996 sum += *sp++;
2997 sum += *sp++;
2998 }
2999 #endif
3000
3001 switch (l4proto)
3002 {
3003 case IPPROTO_UDP :
3004 csump = &((udphdr_t *)l4hdr)->uh_sum;
3005 l4hlen = sizeof(udphdr_t);
3006 break;
3007
3008 case IPPROTO_TCP :
3009 csump = &((tcphdr_t *)l4hdr)->th_sum;
3010 l4hlen = sizeof(tcphdr_t);
3011 break;
3012 case IPPROTO_ICMP :
3013 csump = &((icmphdr_t *)l4hdr)->icmp_cksum;
3014 l4hlen = 4;
3015 sum = 0;
3016 break;
3017 default :
3018 break;
3019 }
3020
3021 if (csump != NULL) {
3022 sumsave = *csump;
3023 *csump = 0;
3024 }
3025
3026 l4hlen = l4hlen; /* LINT */
3027
3028 #ifdef _KERNEL
3029 # ifdef MENTAT
3030 {
3031 void *rp = m->b_rptr;
3032
3033 if ((unsigned char *)ip > m->b_rptr && (unsigned char *)ip < m->b_wptr)
3034 m->b_rptr = (u_char *)ip;
3035 sum2 = ip_cksum(m, hlen, sum); /* hlen == offset */
3036 m->b_rptr = rp;
3037 sum2 = (sum2 & 0xffff) + (sum2 >> 16);
3038 sum2 = ~sum2 & 0xffff;
3039 }
3040 # else /* MENTAT */
3041 # if defined(BSD) || defined(sun)
3042 # if BSD >= 199103
3043 m->m_data += hlen;
3044 # else
3045 m->m_off += hlen;
3046 # endif
3047 m->m_len -= hlen;
3048 sum2 = in_cksum(m, slen);
3049 m->m_len += hlen;
3050 # if BSD >= 199103
3051 m->m_data -= hlen;
3052 # else
3053 m->m_off -= hlen;
3054 # endif
3055 /*
3056 * Both sum and sum2 are partial sums, so combine them together.
3057 */
3058 sum += ~sum2 & 0xffff;
3059 while (sum > 0xffff)
3060 sum = (sum & 0xffff) + (sum >> 16);
3061 sum2 = ~sum & 0xffff;
3062 # else /* defined(BSD) || defined(sun) */
3063 {
3064 union {
3065 u_char c[2];
3066 u_short s;
3067 } bytes;
3068 u_short len = ip->ip_len;
3069 # if defined(__sgi)
3070 int add;
3071 # endif
3072
3073 /*
3074 * Add up IP Header portion
3075 */
3076 if (sp != (u_short *)l4hdr)
3077 sp = (u_short *)l4hdr;
3078
3079 switch (l4proto)
3080 {
3081 case IPPROTO_UDP :
3082 sum += *sp++; /* sport */
3083 sum += *sp++; /* dport */
3084 sum += *sp++; /* udp length */
3085 sum += *sp++; /* checksum */
3086 break;
3087
3088 case IPPROTO_TCP :
3089 sum += *sp++; /* sport */
3090 sum += *sp++; /* dport */
3091 sum += *sp++; /* seq */
3092 sum += *sp++;
3093 sum += *sp++; /* ack */
3094 sum += *sp++;
3095 sum += *sp++; /* off */
3096 sum += *sp++; /* win */
3097 sum += *sp++; /* checksum */
3098 sum += *sp++; /* urp */
3099 break;
3100 case IPPROTO_ICMP :
3101 sum = *sp++; /* type/code */
3102 sum += *sp++; /* checksum */
3103 break;
3104 }
3105
3106 # ifdef __sgi
3107 /*
3108 * In case we had to copy the IP & TCP header out of mbufs,
3109 * skip over the mbuf bits which are the header
3110 */
3111 if ((caddr_t)ip != mtod(m, caddr_t)) {
3112 hlen = (caddr_t)sp - (caddr_t)ip;
3113 while (hlen) {
3114 add = MIN(hlen, m->m_len);
3115 sp = (u_short *)(mtod(m, caddr_t) + add);
3116 hlen -= add;
3117 if (add == m->m_len) {
3118 m = m->m_next;
3119 if (!hlen) {
3120 if (!m)
3121 break;
3122 sp = mtod(m, u_short *);
3123 }
3124 PANIC((!m),("fr_cksum(1): not enough data"));
3125 }
3126 }
3127 }
3128 # endif
3129
3130 len -= (l4hlen + hlen);
3131 if (len <= 0)
3132 goto nodata;
3133
3134 while (len > 1) {
3135 if (((caddr_t)sp - mtod(m, caddr_t)) >= m->m_len) {
3136 m = m->m_next;
3137 PANIC((!m),("fr_cksum(2): not enough data"));
3138 sp = mtod(m, u_short *);
3139 }
3140 if (((caddr_t)(sp + 1) - mtod(m, caddr_t)) > m->m_len) {
3141 bytes.c[0] = *(u_char *)sp;
3142 m = m->m_next;
3143 PANIC((!m),("fr_cksum(3): not enough data"));
3144 sp = mtod(m, u_short *);
3145 bytes.c[1] = *(u_char *)sp;
3146 sum += bytes.s;
3147 sp = (u_short *)((u_char *)sp + 1);
3148 }
3149 if ((u_long)sp & 1) {
3150 bcopy((char *)sp++, (char *)&bytes.s, sizeof(bytes.s));
3151 sum += bytes.s;
3152 } else
3153 sum += *sp++;
3154 len -= 2;
3155 }
3156
3157 if (len != 0)
3158 sum += ntohs(*(u_char *)sp << 8);
3159 nodata:
3160 while (sum > 0xffff)
3161 sum = (sum & 0xffff) + (sum >> 16);
3162 sum2 = (u_short)(~sum & 0xffff);
3163 }
3164 # endif /* defined(BSD) || defined(sun) */
3165 # endif /* MENTAT */
3166 #else /* _KERNEL */
3167 for (; slen > 1; slen -= 2)
3168 sum += *sp++;
3169 if (slen)
3170 sum += ntohs(*(u_char *)sp << 8);
3171 while (sum > 0xffff)
3172 sum = (sum & 0xffff) + (sum >> 16);
3173 sum2 = (u_short)(~sum & 0xffff);
3174 #endif /* _KERNEL */
3175 if (csump != NULL)
3176 *csump = sumsave;
3177 return sum2;
3178 }
3179
3180
3181 #if defined(_KERNEL) && ( ((BSD < 199103) && !defined(MENTAT)) || \
3182 defined(__sgi) ) && !defined(linux) && !defined(_AIX51)
3183 /*
3184 * Copyright (c) 1982, 1986, 1988, 1991, 1993
3185 * The Regents of the University of California. All rights reserved.
3186 *
3187 * Redistribution and use in source and binary forms, with or without
3188 * modification, are permitted provided that the following conditions
3189 * are met:
3190 * 1. Redistributions of source code must retain the above copyright
3191 * notice, this list of conditions and the following disclaimer.
3192 * 2. Redistributions in binary form must reproduce the above copyright
3193 * notice, this list of conditions and the following disclaimer in the
3194 * documentation and/or other materials provided with the distribution.
3195 * 3. Neither the name of the University nor the names of its contributors
3196 * may be used to endorse or promote products derived from this software
3197 * without specific prior written permission.
3198 *
3199 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
3200 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
3201 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
3202 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
3203 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
3204 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
3205 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
3206 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
3207 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
3208 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
3209 * SUCH DAMAGE.
3210 *
3211 * @(#)uipc_mbuf.c 8.2 (Berkeley) 1/4/94
3212 * $Id: fil.c,v 2.243.2.64 2005/08/13 05:19:59 darrenr Exp $
3213 */
3214 /*
3215 * Copy data from an mbuf chain starting "off" bytes from the beginning,
3216 * continuing for "len" bytes, into the indicated buffer.
3217 */
3218 void
m_copydata(m,off,len,cp)3219 m_copydata(m, off, len, cp)
3220 mb_t *m;
3221 int off;
3222 int len;
3223 caddr_t cp;
3224 {
3225 unsigned count;
3226
3227 if (off < 0 || len < 0)
3228 panic("m_copydata");
3229 while (off > 0) {
3230 if (m == 0)
3231 panic("m_copydata");
3232 if (off < m->m_len)
3233 break;
3234 off -= m->m_len;
3235 m = m->m_next;
3236 }
3237 while (len > 0) {
3238 if (m == 0)
3239 panic("m_copydata");
3240 count = MIN(m->m_len - off, len);
3241 bcopy(mtod(m, caddr_t) + off, cp, count);
3242 len -= count;
3243 cp += count;
3244 off = 0;
3245 m = m->m_next;
3246 }
3247 }
3248
3249
3250 /*
3251 * Copy data from a buffer back into the indicated mbuf chain,
3252 * starting "off" bytes from the beginning, extending the mbuf
3253 * chain if necessary.
3254 */
3255 void
m_copyback(m0,off,len,cp)3256 m_copyback(m0, off, len, cp)
3257 struct mbuf *m0;
3258 int off;
3259 int len;
3260 caddr_t cp;
3261 {
3262 int mlen;
3263 struct mbuf *m = m0, *n;
3264 int totlen = 0;
3265
3266 if (m0 == 0)
3267 return;
3268 while (off > (mlen = m->m_len)) {
3269 off -= mlen;
3270 totlen += mlen;
3271 if (m->m_next == 0) {
3272 n = m_getclr(M_DONTWAIT, m->m_type);
3273 if (n == 0)
3274 goto out;
3275 n->m_len = min(MLEN, len + off);
3276 m->m_next = n;
3277 }
3278 m = m->m_next;
3279 }
3280 while (len > 0) {
3281 mlen = min(m->m_len - off, len);
3282 bcopy(cp, off + mtod(m, caddr_t), (unsigned)mlen);
3283 cp += mlen;
3284 len -= mlen;
3285 mlen += off;
3286 off = 0;
3287 totlen += mlen;
3288 if (len == 0)
3289 break;
3290 if (m->m_next == 0) {
3291 n = m_get(M_DONTWAIT, m->m_type);
3292 if (n == 0)
3293 break;
3294 n->m_len = min(MLEN, len);
3295 m->m_next = n;
3296 }
3297 m = m->m_next;
3298 }
3299 out:
3300 #if 0
3301 if (((m = m0)->m_flags & M_PKTHDR) && (m->m_pkthdr.len < totlen))
3302 m->m_pkthdr.len = totlen;
3303 #endif
3304 return;
3305 }
3306 #endif /* (_KERNEL) && ( ((BSD < 199103) && !MENTAT) || __sgi) */
3307
3308
3309 /* ------------------------------------------------------------------------ */
3310 /* Function: fr_findgroup */
3311 /* Returns: frgroup_t * - NULL = group not found, else pointer to group */
3312 /* Parameters: group(I) - group name to search for */
3313 /* unit(I) - device to which this group belongs */
3314 /* set(I) - which set of rules (inactive/inactive) this is */
3315 /* fgpp(O) - pointer to place to store pointer to the pointer */
3316 /* to where to add the next (last) group or where */
3317 /* to delete group from. */
3318 /* */
3319 /* Search amongst the defined groups for a particular group number. */
3320 /* ------------------------------------------------------------------------ */
fr_findgroup(group,unit,set,fgpp,ifs)3321 frgroup_t *fr_findgroup(group, unit, set, fgpp, ifs)
3322 char *group;
3323 minor_t unit;
3324 int set;
3325 frgroup_t ***fgpp;
3326 ipf_stack_t *ifs;
3327 {
3328 frgroup_t *fg, **fgp;
3329
3330 /*
3331 * Which list of groups to search in is dependent on which list of
3332 * rules are being operated on.
3333 */
3334 fgp = &ifs->ifs_ipfgroups[unit][set];
3335
3336 while ((fg = *fgp) != NULL) {
3337 if (strncmp(group, fg->fg_name, FR_GROUPLEN) == 0)
3338 break;
3339 else
3340 fgp = &fg->fg_next;
3341 }
3342 if (fgpp != NULL)
3343 *fgpp = fgp;
3344 return fg;
3345 }
3346
3347
3348 /* ------------------------------------------------------------------------ */
3349 /* Function: fr_addgroup */
3350 /* Returns: frgroup_t * - NULL == did not create group, */
3351 /* != NULL == pointer to the group */
3352 /* Parameters: num(I) - group number to add */
3353 /* head(I) - rule pointer that is using this as the head */
3354 /* flags(I) - rule flags which describe the type of rule it is */
3355 /* unit(I) - device to which this group will belong to */
3356 /* set(I) - which set of rules (inactive/inactive) this is */
3357 /* Write Locks: ipf_mutex */
3358 /* */
3359 /* Add a new group head, or if it already exists, increase the reference */
3360 /* count to it. */
3361 /* ------------------------------------------------------------------------ */
fr_addgroup(group,head,flags,unit,set,ifs)3362 frgroup_t *fr_addgroup(group, head, flags, unit, set, ifs)
3363 char *group;
3364 void *head;
3365 u_32_t flags;
3366 minor_t unit;
3367 int set;
3368 ipf_stack_t *ifs;
3369 {
3370 frgroup_t *fg, **fgp;
3371 u_32_t gflags;
3372
3373 if (group == NULL)
3374 return NULL;
3375
3376 if (unit == IPL_LOGIPF && *group == '\0')
3377 return NULL;
3378
3379 fgp = NULL;
3380 gflags = flags & FR_INOUT;
3381
3382 fg = fr_findgroup(group, unit, set, &fgp, ifs);
3383 if (fg != NULL) {
3384 if (fg->fg_flags == 0)
3385 fg->fg_flags = gflags;
3386 else if (gflags != fg->fg_flags)
3387 return NULL;
3388 fg->fg_ref++;
3389 return fg;
3390 }
3391 KMALLOC(fg, frgroup_t *);
3392 if (fg != NULL) {
3393 fg->fg_head = head;
3394 fg->fg_start = NULL;
3395 fg->fg_next = *fgp;
3396 bcopy(group, fg->fg_name, FR_GROUPLEN);
3397 fg->fg_flags = gflags;
3398 fg->fg_ref = 1;
3399 *fgp = fg;
3400 }
3401 return fg;
3402 }
3403
3404
3405 /* ------------------------------------------------------------------------ */
3406 /* Function: fr_delgroup */
3407 /* Returns: Nil */
3408 /* Parameters: group(I) - group name to delete */
3409 /* unit(I) - device to which this group belongs */
3410 /* set(I) - which set of rules (inactive/inactive) this is */
3411 /* Write Locks: ipf_mutex */
3412 /* */
3413 /* Attempt to delete a group head. */
3414 /* Only do this when its reference count reaches 0. */
3415 /* ------------------------------------------------------------------------ */
fr_delgroup(group,unit,set,ifs)3416 void fr_delgroup(group, unit, set, ifs)
3417 char *group;
3418 minor_t unit;
3419 int set;
3420 ipf_stack_t *ifs;
3421 {
3422 frgroup_t *fg, **fgp;
3423
3424 fg = fr_findgroup(group, unit, set, &fgp, ifs);
3425 if (fg == NULL)
3426 return;
3427
3428 fg->fg_ref--;
3429 if (fg->fg_ref == 0) {
3430 *fgp = fg->fg_next;
3431 KFREE(fg);
3432 }
3433 }
3434
3435
3436 /* ------------------------------------------------------------------------ */
3437 /* Function: fr_getrulen */
3438 /* Returns: frentry_t * - NULL == not found, else pointer to rule n */
3439 /* Parameters: unit(I) - device for which to count the rule's number */
3440 /* flags(I) - which set of rules to find the rule in */
3441 /* group(I) - group name */
3442 /* n(I) - rule number to find */
3443 /* */
3444 /* Find rule # n in group # g and return a pointer to it. Return NULl if */
3445 /* group # g doesn't exist or there are less than n rules in the group. */
3446 /* ------------------------------------------------------------------------ */
fr_getrulen(unit,group,n,ifs)3447 frentry_t *fr_getrulen(unit, group, n, ifs)
3448 int unit;
3449 char *group;
3450 u_32_t n;
3451 ipf_stack_t *ifs;
3452 {
3453 frentry_t *fr;
3454 frgroup_t *fg;
3455
3456 fg = fr_findgroup(group, unit, ifs->ifs_fr_active, NULL, ifs);
3457 if (fg == NULL)
3458 return NULL;
3459 for (fr = fg->fg_head; fr && n; fr = fr->fr_next, n--)
3460 ;
3461 if (n != 0)
3462 return NULL;
3463 return fr;
3464 }
3465
3466
3467 /* ------------------------------------------------------------------------ */
3468 /* Function: fr_rulen */
3469 /* Returns: int - >= 0 - rule number, -1 == search failed */
3470 /* Parameters: unit(I) - device for which to count the rule's number */
3471 /* fr(I) - pointer to rule to match */
3472 /* */
3473 /* Return the number for a rule on a specific filtering device. */
3474 /* ------------------------------------------------------------------------ */
fr_rulen(unit,fr,ifs)3475 int fr_rulen(unit, fr, ifs)
3476 int unit;
3477 frentry_t *fr;
3478 ipf_stack_t *ifs;
3479 {
3480 frentry_t *fh;
3481 frgroup_t *fg;
3482 u_32_t n = 0;
3483
3484 if (fr == NULL)
3485 return -1;
3486 fg = fr_findgroup(fr->fr_group, unit, ifs->ifs_fr_active, NULL, ifs);
3487 if (fg == NULL)
3488 return -1;
3489 for (fh = fg->fg_head; fh; n++, fh = fh->fr_next)
3490 if (fh == fr)
3491 break;
3492 if (fh == NULL)
3493 return -1;
3494 return n;
3495 }
3496
3497
3498 /* ------------------------------------------------------------------------ */
3499 /* Function: frflushlist */
3500 /* Returns: int - >= 0 - number of flushed rules */
3501 /* Parameters: set(I) - which set of rules (inactive/inactive) this is */
3502 /* unit(I) - device for which to flush rules */
3503 /* flags(I) - which set of rules to flush */
3504 /* nfreedp(O) - pointer to int where flush count is stored */
3505 /* listp(I) - pointer to list to flush pointer */
3506 /* Write Locks: ipf_mutex */
3507 /* */
3508 /* Recursively flush rules from the list, descending groups as they are */
3509 /* encountered. if a rule is the head of a group and it has lost all its */
3510 /* group members, then also delete the group reference. nfreedp is needed */
3511 /* to store the accumulating count of rules removed, whereas the returned */
3512 /* value is just the number removed from the current list. The latter is */
3513 /* needed to correctly adjust reference counts on rules that define groups. */
3514 /* */
3515 /* NOTE: Rules not loaded from user space cannot be flushed. */
3516 /* ------------------------------------------------------------------------ */
frflushlist(set,unit,nfreedp,listp,ifs)3517 static int frflushlist(set, unit, nfreedp, listp, ifs)
3518 int set;
3519 minor_t unit;
3520 int *nfreedp;
3521 frentry_t **listp;
3522 ipf_stack_t *ifs;
3523 {
3524 int freed = 0;
3525 frentry_t *fp;
3526
3527 while ((fp = *listp) != NULL) {
3528 if ((fp->fr_type & FR_T_BUILTIN) ||
3529 !(fp->fr_flags & FR_COPIED)) {
3530 listp = &fp->fr_next;
3531 continue;
3532 }
3533 *listp = fp->fr_next;
3534 if (fp->fr_grp != NULL) {
3535 (void) frflushlist(set, unit, nfreedp, fp->fr_grp, ifs);
3536 }
3537
3538 fr_delgroup(fp->fr_grhead, unit, set, ifs);
3539 *fp->fr_grhead = '\0';
3540
3541 ASSERT(fp->fr_ref > 0);
3542 fp->fr_next = NULL;
3543 if (fr_derefrule(&fp, ifs) == 0)
3544 freed++;
3545 }
3546 *nfreedp += freed;
3547 return freed;
3548 }
3549
3550
3551 /* ------------------------------------------------------------------------ */
3552 /* Function: frflush */
3553 /* Returns: int - >= 0 - number of flushed rules */
3554 /* Parameters: unit(I) - device for which to flush rules */
3555 /* flags(I) - which set of rules to flush */
3556 /* */
3557 /* Calls flushlist() for all filter rules (accounting, firewall - both IPv4 */
3558 /* and IPv6) as defined by the value of flags. */
3559 /* ------------------------------------------------------------------------ */
frflush(unit,proto,flags,ifs)3560 int frflush(unit, proto, flags, ifs)
3561 minor_t unit;
3562 int proto, flags;
3563 ipf_stack_t *ifs;
3564 {
3565 int flushed = 0, set;
3566
3567 WRITE_ENTER(&ifs->ifs_ipf_mutex);
3568 bzero((char *)ifs->ifs_frcache, sizeof (ifs->ifs_frcache));
3569
3570 set = ifs->ifs_fr_active;
3571 if ((flags & FR_INACTIVE) == FR_INACTIVE)
3572 set = 1 - set;
3573
3574 if (flags & FR_OUTQUE) {
3575 if (proto == 0 || proto == 6) {
3576 (void) frflushlist(set, unit,
3577 &flushed, &ifs->ifs_ipfilter6[1][set], ifs);
3578 (void) frflushlist(set, unit,
3579 &flushed, &ifs->ifs_ipacct6[1][set], ifs);
3580 }
3581 if (proto == 0 || proto == 4) {
3582 (void) frflushlist(set, unit,
3583 &flushed, &ifs->ifs_ipfilter[1][set], ifs);
3584 (void) frflushlist(set, unit,
3585 &flushed, &ifs->ifs_ipacct[1][set], ifs);
3586 }
3587 }
3588 if (flags & FR_INQUE) {
3589 if (proto == 0 || proto == 6) {
3590 (void) frflushlist(set, unit,
3591 &flushed, &ifs->ifs_ipfilter6[0][set], ifs);
3592 (void) frflushlist(set, unit,
3593 &flushed, &ifs->ifs_ipacct6[0][set], ifs);
3594 }
3595 if (proto == 0 || proto == 4) {
3596 (void) frflushlist(set, unit,
3597 &flushed, &ifs->ifs_ipfilter[0][set], ifs);
3598 (void) frflushlist(set, unit,
3599 &flushed, &ifs->ifs_ipacct[0][set], ifs);
3600 }
3601 }
3602 RWLOCK_EXIT(&ifs->ifs_ipf_mutex);
3603
3604 if (unit == IPL_LOGIPF) {
3605 int tmp;
3606
3607 tmp = frflush(IPL_LOGCOUNT, proto, flags, ifs);
3608 if (tmp >= 0)
3609 flushed += tmp;
3610 }
3611 return flushed;
3612 }
3613
3614
3615 /* ------------------------------------------------------------------------ */
3616 /* Function: memstr */
3617 /* Returns: char * - NULL if failed, != NULL pointer to matching bytes */
3618 /* Parameters: src(I) - pointer to byte sequence to match */
3619 /* dst(I) - pointer to byte sequence to search */
3620 /* slen(I) - match length */
3621 /* dlen(I) - length available to search in */
3622 /* */
3623 /* Search dst for a sequence of bytes matching those at src and extend for */
3624 /* slen bytes. */
3625 /* ------------------------------------------------------------------------ */
memstr(src,dst,slen,dlen)3626 char *memstr(src, dst, slen, dlen)
3627 char *src, *dst;
3628 int slen, dlen;
3629 {
3630 char *s = NULL;
3631
3632 while (dlen >= slen) {
3633 if (bcmp(src, dst, slen) == 0) {
3634 s = dst;
3635 break;
3636 }
3637 dst++;
3638 dlen--;
3639 }
3640 return s;
3641 }
3642 /* ------------------------------------------------------------------------ */
3643 /* Function: fr_fixskip */
3644 /* Returns: Nil */
3645 /* Parameters: listp(IO) - pointer to start of list with skip rule */
3646 /* rp(I) - rule added/removed with skip in it. */
3647 /* addremove(I) - adjustment (-1/+1) to make to skip count, */
3648 /* depending on whether a rule was just added */
3649 /* or removed. */
3650 /* */
3651 /* Adjust all the rules in a list which would have skip'd past the position */
3652 /* where we are inserting to skip to the right place given the change. */
3653 /* ------------------------------------------------------------------------ */
fr_fixskip(listp,rp,addremove)3654 void fr_fixskip(listp, rp, addremove)
3655 frentry_t **listp, *rp;
3656 int addremove;
3657 {
3658 int rules, rn;
3659 frentry_t *fp;
3660
3661 rules = 0;
3662 for (fp = *listp; (fp != NULL) && (fp != rp); fp = fp->fr_next)
3663 rules++;
3664
3665 if (!fp)
3666 return;
3667
3668 for (rn = 0, fp = *listp; fp && (fp != rp); fp = fp->fr_next, rn++)
3669 if (FR_ISSKIP(fp->fr_flags) && (rn + fp->fr_arg >= rules))
3670 fp->fr_arg += addremove;
3671 }
3672
3673
3674 #ifdef _KERNEL
3675 /* ------------------------------------------------------------------------ */
3676 /* Function: count4bits */
3677 /* Returns: int - >= 0 - number of consecutive bits in input */
3678 /* Parameters: ip(I) - 32bit IP address */
3679 /* */
3680 /* IPv4 ONLY */
3681 /* count consecutive 1's in bit mask. If the mask generated by counting */
3682 /* consecutive 1's is different to that passed, return -1, else return # */
3683 /* of bits. */
3684 /* ------------------------------------------------------------------------ */
count4bits(ip)3685 int count4bits(ip)
3686 u_32_t ip;
3687 {
3688 u_32_t ipn;
3689 int cnt = 0, i, j;
3690
3691 ip = ipn = ntohl(ip);
3692 for (i = 32; i; i--, ipn *= 2)
3693 if (ipn & 0x80000000)
3694 cnt++;
3695 else
3696 break;
3697 ipn = 0;
3698 for (i = 32, j = cnt; i; i--, j--) {
3699 ipn *= 2;
3700 if (j > 0)
3701 ipn++;
3702 }
3703 if (ipn == ip)
3704 return cnt;
3705 return -1;
3706 }
3707
3708
3709 #ifdef USE_INET6
3710 /* ------------------------------------------------------------------------ */
3711 /* Function: count6bits */
3712 /* Returns: int - >= 0 - number of consecutive bits in input */
3713 /* Parameters: msk(I) - pointer to start of IPv6 bitmask */
3714 /* */
3715 /* IPv6 ONLY */
3716 /* count consecutive 1's in bit mask. */
3717 /* ------------------------------------------------------------------------ */
count6bits(msk)3718 int count6bits(msk)
3719 u_32_t *msk;
3720 {
3721 int i = 0, k;
3722 u_32_t j;
3723
3724 for (k = 3; k >= 0; k--)
3725 if (msk[k] == 0xffffffff)
3726 i += 32;
3727 else {
3728 for (j = msk[k]; j; j <<= 1)
3729 if (j & 0x80000000)
3730 i++;
3731 }
3732 return i;
3733 }
3734 # endif
3735 #endif /* _KERNEL */
3736
3737
3738 /* ------------------------------------------------------------------------ */
3739 /* Function: fr_ifsync */
3740 /* Returns: void * - new interface identifier */
3741 /* Parameters: action(I) - type of synchronisation to do */
3742 /* v(I) - IP version being sync'd (v4 or v6) */
3743 /* newifp(I) - interface identifier being introduced/removed */
3744 /* oldifp(I) - interface identifier in a filter rule */
3745 /* newname(I) - name associated with newifp interface */
3746 /* oldname(I) - name associated with oldifp interface */
3747 /* ifs - pointer to IPF stack instance */
3748 /* */
3749 /* This function returns what the new value for "oldifp" should be for its */
3750 /* caller. In some cases it will not change, in some it will. */
3751 /* action == IPFSYNC_RESYNC */
3752 /* a new value for oldifp will always be looked up, according to oldname, */
3753 /* the values of newname and newifp are ignored. */
3754 /* action == IPFSYNC_NEWIFP */
3755 /* if oldname matches newname then we are doing a sync for the matching */
3756 /* interface, so we return newifp to be used in place of oldifp. If the */
3757 /* the names don't match, just return oldifp. */
3758 /* action == IPFSYNC_OLDIFP */
3759 /* if oldifp matches newifp then we are are doing a sync to remove any */
3760 /* references to oldifp, so we return "-1". */
3761 /* ----- */
3762 /* NOTE: */
3763 /* This function processes NIC event from PF_HOOKS. The action parameter */
3764 /* is set in ipf_nic_event_v4()/ipf_nic_event_v6() function. There is */
3765 /* one single switch statement() in ipf_nic_event_vx() function, which */
3766 /* translates the HOOK event type to action parameter passed to fr_ifsync. */
3767 /* The translation table looks as follows: */
3768 /* event | action */
3769 /* ----------------+------------- */
3770 /* NE_PLUMB | IPFSYNC_NEWIFP */
3771 /* NE_UNPLUMB | IPFSYNC_OLDIFP */
3772 /* NE_ADDRESS_CHANGE | IPFSYNC_RESYNC */
3773 /* */
3774 /* The oldname and oldifp parameters are taken from IPF entry (rule, state */
3775 /* table entry, NAT table entry, fragment ...). The newname and newifp */
3776 /* parameters come from hook event data, parameters are taken from event */
3777 /* in ipf_nic_event_vx() functions. Any time NIC changes, the IPF is */
3778 /* notified by hook function. */
3779 /* */
3780 /* We get NE_UNPLUMB event from PF_HOOKS even if someone coincidently tries */
3781 /* to plumb the interface, which is already plumbed. In such case we always */
3782 /* get the event from PF_HOOKS as follows: */
3783 /* event: NE_PLUMB */
3784 /* NIC: 0x0 */
3785 /* ------------------------------------------------------------------------ */
fr_ifsync(action,v,newname,oldname,newifp,oldifp,ifs)3786 static void *fr_ifsync(action, v, newname, oldname, newifp, oldifp, ifs)
3787 int action, v;
3788 char *newname, *oldname;
3789 void *newifp, *oldifp;
3790 ipf_stack_t *ifs;
3791 {
3792 void *rval = oldifp;
3793
3794 switch (action)
3795 {
3796 case IPFSYNC_RESYNC :
3797 if (oldname[0] != '\0') {
3798 rval = fr_resolvenic(oldname, v, ifs);
3799 }
3800 break;
3801 case IPFSYNC_NEWIFP :
3802 if (!strncmp(newname, oldname, LIFNAMSIZ))
3803 rval = newifp;
3804 break;
3805 case IPFSYNC_OLDIFP :
3806 /*
3807 * If interface gets unplumbed it must be invalidated, which
3808 * means set all existing references to the interface to -1.
3809 * We don't want to invalidate references for wildcard
3810 * (unbound) rules (entries).
3811 */
3812 if (newifp == oldifp)
3813 rval = (oldifp) ? (void *)-1 : NULL;
3814 break;
3815 }
3816
3817 return rval;
3818 }
3819
3820
3821 /* ------------------------------------------------------------------------ */
3822 /* Function: frsynclist */
3823 /* Returns: void */
3824 /* Parameters: action(I) - type of synchronisation to do */
3825 /* v(I) - IP version being sync'd (v4 or v6) */
3826 /* ifp(I) - interface identifier associated with action */
3827 /* ifname(I) - name associated with ifp parameter */
3828 /* fr(I) - pointer to filter rule */
3829 /* ifs - pointer to IPF stack instance */
3830 /* Write Locks: ipf_mutex */
3831 /* */
3832 /* Walk through a list of filter rules and resolve any interface names into */
3833 /* pointers. Where dynamic addresses are used, also update the IP address */
3834 /* used in the rule. The interface pointer is used to limit the lookups to */
3835 /* a specific set of matching names if it is non-NULL. */
3836 /* ------------------------------------------------------------------------ */
frsynclist(action,v,ifp,ifname,fr,ifs)3837 static void frsynclist(action, v, ifp, ifname, fr, ifs)
3838 int action, v;
3839 void *ifp;
3840 char *ifname;
3841 frentry_t *fr;
3842 ipf_stack_t *ifs;
3843 {
3844 frdest_t *fdp;
3845 int rv, i;
3846
3847 for (; fr; fr = fr->fr_next) {
3848 rv = fr->fr_v;
3849 if (v != 0 && v != rv)
3850 continue;
3851
3852 /*
3853 * Lookup all the interface names that are part of the rule.
3854 */
3855 for (i = 0; i < 4; i++) {
3856 fr->fr_ifas[i] = fr_ifsync(action, rv, ifname,
3857 fr->fr_ifnames[i],
3858 ifp, fr->fr_ifas[i],
3859 ifs);
3860 }
3861
3862 fdp = &fr->fr_tifs[0];
3863 fdp->fd_ifp = fr_ifsync(action, rv, ifname, fdp->fd_ifname,
3864 ifp, fdp->fd_ifp, ifs);
3865
3866 fdp = &fr->fr_tifs[1];
3867 fdp->fd_ifp = fr_ifsync(action, rv, ifname, fdp->fd_ifname,
3868 ifp, fdp->fd_ifp, ifs);
3869
3870 fdp = &fr->fr_dif;
3871 fdp->fd_ifp = fr_ifsync(action, rv, ifname, fdp->fd_ifname,
3872 ifp, fdp->fd_ifp, ifs);
3873
3874 if (action != IPFSYNC_RESYNC)
3875 continue;
3876
3877 if (fr->fr_type == FR_T_IPF) {
3878 if (fr->fr_satype != FRI_NORMAL &&
3879 fr->fr_satype != FRI_LOOKUP) {
3880 (void)fr_ifpaddr(rv, fr->fr_satype,
3881 fr->fr_ifas[fr->fr_sifpidx],
3882 &fr->fr_src, &fr->fr_smsk,
3883 ifs);
3884 }
3885 if (fr->fr_datype != FRI_NORMAL &&
3886 fr->fr_datype != FRI_LOOKUP) {
3887 (void)fr_ifpaddr(rv, fr->fr_datype,
3888 fr->fr_ifas[fr->fr_difpidx],
3889 &fr->fr_dst, &fr->fr_dmsk,
3890 ifs);
3891 }
3892 }
3893
3894 #ifdef IPFILTER_LOOKUP
3895 if (fr->fr_type == FR_T_IPF && fr->fr_satype == FRI_LOOKUP &&
3896 fr->fr_srcptr == NULL) {
3897 fr->fr_srcptr = fr_resolvelookup(fr->fr_srctype,
3898 fr->fr_srcnum,
3899 &fr->fr_srcfunc, ifs);
3900 }
3901 if (fr->fr_type == FR_T_IPF && fr->fr_datype == FRI_LOOKUP &&
3902 fr->fr_dstptr == NULL) {
3903 fr->fr_dstptr = fr_resolvelookup(fr->fr_dsttype,
3904 fr->fr_dstnum,
3905 &fr->fr_dstfunc, ifs);
3906 }
3907 #endif
3908 }
3909 }
3910
3911
3912 #ifdef _KERNEL
3913 /* ------------------------------------------------------------------------ */
3914 /* Function: frsync */
3915 /* Returns: void */
3916 /* Parameters: action(I) - type of synchronisation to do */
3917 /* v(I) - IP version being sync'd (v4 or v6) */
3918 /* ifp(I) - interface identifier associated with action */
3919 /* name(I) - name associated with ifp parameter */
3920 /* */
3921 /* frsync() is called when we suspect that the interface list or */
3922 /* information about interfaces (like IP#) has changed. Go through all */
3923 /* filter rules, NAT entries and the state table and check if anything */
3924 /* needs to be changed/updated. */
3925 /* With the filtering hooks added to Solaris, we needed to change the manner*/
3926 /* in which this was done to support three different types of sync: */
3927 /* - complete resync of all interface name/identifiers */
3928 /* - new interface being announced with its name and identifier */
3929 /* - interface removal being announced by only its identifier */
3930 /* ------------------------------------------------------------------------ */
frsync(action,v,ifp,name,ifs)3931 void frsync(action, v, ifp, name, ifs)
3932 int action, v;
3933 void *ifp;
3934 char *name;
3935 ipf_stack_t *ifs;
3936 {
3937 int i;
3938
3939 WRITE_ENTER(&ifs->ifs_ipf_mutex);
3940 frsynclist(action, v, ifp, name, ifs->ifs_ipacct[0][ifs->ifs_fr_active], ifs);
3941 frsynclist(action, v, ifp, name, ifs->ifs_ipacct[1][ifs->ifs_fr_active], ifs);
3942 frsynclist(action, v, ifp, name, ifs->ifs_ipfilter[0][ifs->ifs_fr_active], ifs);
3943 frsynclist(action, v, ifp, name, ifs->ifs_ipfilter[1][ifs->ifs_fr_active], ifs);
3944 frsynclist(action, v, ifp, name, ifs->ifs_ipacct6[0][ifs->ifs_fr_active], ifs);
3945 frsynclist(action, v, ifp, name, ifs->ifs_ipacct6[1][ifs->ifs_fr_active], ifs);
3946 frsynclist(action, v, ifp, name, ifs->ifs_ipfilter6[0][ifs->ifs_fr_active], ifs);
3947 frsynclist(action, v, ifp, name, ifs->ifs_ipfilter6[1][ifs->ifs_fr_active], ifs);
3948
3949 for (i = 0; i < IPL_LOGSIZE; i++) {
3950 frgroup_t *g;
3951
3952 for (g = ifs->ifs_ipfgroups[i][0]; g != NULL; g = g->fg_next)
3953 frsynclist(action, v, ifp, name, g->fg_start, ifs);
3954 for (g = ifs->ifs_ipfgroups[i][1]; g != NULL; g = g->fg_next)
3955 frsynclist(action, v, ifp, name, g->fg_start, ifs);
3956 }
3957 RWLOCK_EXIT(&ifs->ifs_ipf_mutex);
3958 }
3959
3960 #if SOLARIS2 >= 10
3961 /* ------------------------------------------------------------------------ */
3962 /* Function: fr_syncindex */
3963 /* Returns: void */
3964 /* Parameters: rules - list of rules to be sync'd */
3965 /* ifp - interface, which is being sync'd */
3966 /* newifp - new ifindex value for interface */
3967 /* */
3968 /* Function updates all NIC indecis, which match ifp, in every rule. Every */
3969 /* NIC index matching ifp, will be updated to newifp. */
3970 /* ------------------------------------------------------------------------ */
fr_syncindex(rules,ifp,newifp)3971 static void fr_syncindex(rules, ifp, newifp)
3972 frentry_t *rules;
3973 void *ifp;
3974 void *newifp;
3975 {
3976 int i;
3977 frentry_t *fr;
3978
3979 for (fr = rules; fr != NULL; fr = fr->fr_next) {
3980 /*
3981 * Lookup all the interface names that are part of the rule.
3982 */
3983 for (i = 0; i < 4; i++)
3984 if (fr->fr_ifas[i] == ifp)
3985 fr->fr_ifas[i] = newifp;
3986
3987 for (i = 0; i < 2; i++) {
3988 if (fr->fr_tifs[i].fd_ifp == ifp)
3989 fr->fr_tifs[i].fd_ifp = newifp;
3990 }
3991
3992 if (fr->fr_dif.fd_ifp == ifp)
3993 fr->fr_dif.fd_ifp = newifp;
3994 }
3995 }
3996
3997 /* ------------------------------------------------------------------------ */
3998 /* Function: fr_ifindexsync */
3999 /* Returns: void */
4000 /* Parameters: ifp - interface, which is being sync'd */
4001 /* newifp - new ifindex value for interface */
4002 /* ifs - IPF's stack */
4003 /* */
4004 /* Function assumes ipf_mutex is locked exclusively. */
4005 /* */
4006 /* Function updates the NIC references in rules with new interfaces index */
4007 /* (newifp). Function must process active lists: */
4008 /* with accounting rules (IPv6 and IPv4) */
4009 /* with inbound rules (IPv6 and IPv4) */
4010 /* with outbound rules (IPv6 and IPv4) */
4011 /* Function also has to take care of rule groups. */
4012 /* */
4013 /* NOTE: The ipf_mutex is grabbed exclusively by caller (which is always */
4014 /* nic_event_hook). The hook function also updates state entries, NAT rules */
4015 /* and NAT entries. We want to do all these update atomically to keep the */
4016 /* NIC references consistent. The ipf_mutex will synchronize event with */
4017 /* fr_check(), which processes packets, so no packet will enter fr_check(), */
4018 /* while NIC references will be synchronized. */
4019 /* ------------------------------------------------------------------------ */
fr_ifindexsync(ifp,newifp,ifs)4020 void fr_ifindexsync(ifp, newifp, ifs)
4021 void *ifp;
4022 void *newifp;
4023 ipf_stack_t *ifs;
4024 {
4025 unsigned int i;
4026 frentry_t *rule_lists[8];
4027 unsigned int rules = sizeof (rule_lists) / sizeof (frentry_t *);
4028
4029 rule_lists[0] = ifs->ifs_ipacct[0][ifs->ifs_fr_active];
4030 rule_lists[1] = ifs->ifs_ipacct[1][ifs->ifs_fr_active];
4031 rule_lists[2] = ifs->ifs_ipfilter[0][ifs->ifs_fr_active];
4032 rule_lists[3] = ifs->ifs_ipfilter[1][ifs->ifs_fr_active];
4033 rule_lists[4] = ifs->ifs_ipacct6[0][ifs->ifs_fr_active];
4034 rule_lists[5] = ifs->ifs_ipacct6[1][ifs->ifs_fr_active];
4035 rule_lists[6] = ifs->ifs_ipfilter6[0][ifs->ifs_fr_active];
4036 rule_lists[7] = ifs->ifs_ipfilter6[1][ifs->ifs_fr_active];
4037
4038 for (i = 0; i < rules; i++) {
4039 fr_syncindex(rule_lists[i], ifp, newifp);
4040 }
4041
4042 /*
4043 * Update rule groups.
4044 */
4045 for (i = 0; i < IPL_LOGSIZE; i++) {
4046 frgroup_t *g;
4047
4048 for (g = ifs->ifs_ipfgroups[i][0]; g != NULL; g = g->fg_next)
4049 fr_syncindex(g->fg_start, ifp, newifp);
4050 for (g = ifs->ifs_ipfgroups[i][1]; g != NULL; g = g->fg_next)
4051 fr_syncindex(g->fg_start, ifp, newifp);
4052 }
4053 }
4054 #endif
4055
4056 /*
4057 * In the functions below, bcopy() is called because the pointer being
4058 * copied _from_ in this instance is a pointer to a char buf (which could
4059 * end up being unaligned) and on the kernel's local stack.
4060 */
4061 /* ------------------------------------------------------------------------ */
4062 /* Function: copyinptr */
4063 /* Returns: int - 0 = success, else failure */
4064 /* Parameters: src(I) - pointer to the source address */
4065 /* dst(I) - destination address */
4066 /* size(I) - number of bytes to copy */
4067 /* */
4068 /* Copy a block of data in from user space, given a pointer to the pointer */
4069 /* to start copying from (src) and a pointer to where to store it (dst). */
4070 /* NB: src - pointer to user space pointer, dst - kernel space pointer */
4071 /* ------------------------------------------------------------------------ */
copyinptr(src,dst,size)4072 int copyinptr(src, dst, size)
4073 void *src, *dst;
4074 size_t size;
4075 {
4076 caddr_t ca;
4077 int err;
4078
4079 # ifdef SOLARIS
4080 err = COPYIN(src, (caddr_t)&ca, sizeof(ca));
4081 if (err != 0)
4082 return err;
4083 # else
4084 bcopy(src, (caddr_t)&ca, sizeof(ca));
4085 # endif
4086 err = COPYIN(ca, dst, size);
4087 return err;
4088 }
4089
4090
4091 /* ------------------------------------------------------------------------ */
4092 /* Function: copyoutptr */
4093 /* Returns: int - 0 = success, else failure */
4094 /* Parameters: src(I) - pointer to the source address */
4095 /* dst(I) - destination address */
4096 /* size(I) - number of bytes to copy */
4097 /* */
4098 /* Copy a block of data out to user space, given a pointer to the pointer */
4099 /* to start copying from (src) and a pointer to where to store it (dst). */
4100 /* NB: src - kernel space pointer, dst - pointer to user space pointer. */
4101 /* ------------------------------------------------------------------------ */
copyoutptr(src,dst,size)4102 int copyoutptr(src, dst, size)
4103 void *src, *dst;
4104 size_t size;
4105 {
4106 caddr_t ca;
4107 int err;
4108
4109 # ifdef SOLARIS
4110 err = COPYIN(dst, (caddr_t)&ca, sizeof(ca));
4111 if (err != 0)
4112 return err;
4113 # else
4114 bcopy(dst, (caddr_t)&ca, sizeof(ca));
4115 # endif
4116 err = COPYOUT(src, ca, size);
4117 return err;
4118 }
4119 #endif
4120
4121
4122 /* ------------------------------------------------------------------------ */
4123 /* Function: fr_lock */
4124 /* Returns: int - 0 = success, else error */
4125 /* Parameters: data(I) - pointer to lock value to set */
4126 /* lockp(O) - pointer to location to store old lock value */
4127 /* */
4128 /* Get the new value for the lock integer, set it and return the old value */
4129 /* in *lockp. */
4130 /* ------------------------------------------------------------------------ */
fr_lock(data,lockp)4131 int fr_lock(data, lockp)
4132 caddr_t data;
4133 int *lockp;
4134 {
4135 int arg, err;
4136
4137 err = BCOPYIN(data, (caddr_t)&arg, sizeof(arg));
4138 if (err != 0)
4139 return (EFAULT);
4140 err = BCOPYOUT((caddr_t)lockp, data, sizeof(*lockp));
4141 if (err != 0)
4142 return (EFAULT);
4143 *lockp = arg;
4144 return (0);
4145 }
4146
4147
4148 /* ------------------------------------------------------------------------ */
4149 /* Function: fr_getstat */
4150 /* Returns: Nil */
4151 /* Parameters: fiop(I) - pointer to ipfilter stats structure */
4152 /* */
4153 /* Stores a copy of current pointers, counters, etc, in the friostat */
4154 /* structure. */
4155 /* ------------------------------------------------------------------------ */
fr_getstat(fiop,ifs)4156 void fr_getstat(fiop, ifs)
4157 friostat_t *fiop;
4158 ipf_stack_t *ifs;
4159 {
4160 int i, j;
4161
4162 bcopy((char *)&ifs->ifs_frstats, (char *)fiop->f_st,
4163 sizeof(filterstats_t) * 2);
4164 fiop->f_locks[IPL_LOGSTATE] = ifs->ifs_fr_state_lock;
4165 fiop->f_locks[IPL_LOGNAT] = ifs->ifs_fr_nat_lock;
4166 fiop->f_locks[IPL_LOGIPF] = ifs->ifs_fr_frag_lock;
4167 fiop->f_locks[IPL_LOGAUTH] = ifs->ifs_fr_auth_lock;
4168
4169 for (i = 0; i < 2; i++)
4170 for (j = 0; j < 2; j++) {
4171 fiop->f_ipf[i][j] = ifs->ifs_ipfilter[i][j];
4172 fiop->f_acct[i][j] = ifs->ifs_ipacct[i][j];
4173 fiop->f_ipf6[i][j] = ifs->ifs_ipfilter6[i][j];
4174 fiop->f_acct6[i][j] = ifs->ifs_ipacct6[i][j];
4175 }
4176
4177 fiop->f_ticks = ifs->ifs_fr_ticks;
4178 fiop->f_active = ifs->ifs_fr_active;
4179 fiop->f_froute[0] = ifs->ifs_fr_frouteok[0];
4180 fiop->f_froute[1] = ifs->ifs_fr_frouteok[1];
4181
4182 fiop->f_running = ifs->ifs_fr_running;
4183 for (i = 0; i < IPL_LOGSIZE; i++) {
4184 fiop->f_groups[i][0] = ifs->ifs_ipfgroups[i][0];
4185 fiop->f_groups[i][1] = ifs->ifs_ipfgroups[i][1];
4186 }
4187 #ifdef IPFILTER_LOG
4188 fiop->f_logging = 1;
4189 #else
4190 fiop->f_logging = 0;
4191 #endif
4192 fiop->f_defpass = ifs->ifs_fr_pass;
4193 fiop->f_features = fr_features;
4194 (void) strncpy(fiop->f_version, ipfilter_version,
4195 sizeof(fiop->f_version));
4196 }
4197
4198
4199 #ifdef USE_INET6
4200 int icmptoicmp6types[ICMP_MAXTYPE+1] = {
4201 ICMP6_ECHO_REPLY, /* 0: ICMP_ECHOREPLY */
4202 -1, /* 1: UNUSED */
4203 -1, /* 2: UNUSED */
4204 ICMP6_DST_UNREACH, /* 3: ICMP_UNREACH */
4205 -1, /* 4: ICMP_SOURCEQUENCH */
4206 ND_REDIRECT, /* 5: ICMP_REDIRECT */
4207 -1, /* 6: UNUSED */
4208 -1, /* 7: UNUSED */
4209 ICMP6_ECHO_REQUEST, /* 8: ICMP_ECHO */
4210 -1, /* 9: UNUSED */
4211 -1, /* 10: UNUSED */
4212 ICMP6_TIME_EXCEEDED, /* 11: ICMP_TIMXCEED */
4213 ICMP6_PARAM_PROB, /* 12: ICMP_PARAMPROB */
4214 -1, /* 13: ICMP_TSTAMP */
4215 -1, /* 14: ICMP_TSTAMPREPLY */
4216 -1, /* 15: ICMP_IREQ */
4217 -1, /* 16: ICMP_IREQREPLY */
4218 -1, /* 17: ICMP_MASKREQ */
4219 -1, /* 18: ICMP_MASKREPLY */
4220 };
4221
4222
4223 int icmptoicmp6unreach[ICMP_MAX_UNREACH] = {
4224 ICMP6_DST_UNREACH_ADDR, /* 0: ICMP_UNREACH_NET */
4225 ICMP6_DST_UNREACH_ADDR, /* 1: ICMP_UNREACH_HOST */
4226 -1, /* 2: ICMP_UNREACH_PROTOCOL */
4227 ICMP6_DST_UNREACH_NOPORT, /* 3: ICMP_UNREACH_PORT */
4228 -1, /* 4: ICMP_UNREACH_NEEDFRAG */
4229 ICMP6_DST_UNREACH_NOTNEIGHBOR, /* 5: ICMP_UNREACH_SRCFAIL */
4230 ICMP6_DST_UNREACH_ADDR, /* 6: ICMP_UNREACH_NET_UNKNOWN */
4231 ICMP6_DST_UNREACH_ADDR, /* 7: ICMP_UNREACH_HOST_UNKNOWN */
4232 -1, /* 8: ICMP_UNREACH_ISOLATED */
4233 ICMP6_DST_UNREACH_ADMIN, /* 9: ICMP_UNREACH_NET_PROHIB */
4234 ICMP6_DST_UNREACH_ADMIN, /* 10: ICMP_UNREACH_HOST_PROHIB */
4235 -1, /* 11: ICMP_UNREACH_TOSNET */
4236 -1, /* 12: ICMP_UNREACH_TOSHOST */
4237 ICMP6_DST_UNREACH_ADMIN, /* 13: ICMP_UNREACH_ADMIN_PROHIBIT */
4238 };
4239 int icmpreplytype6[ICMP6_MAXTYPE + 1];
4240 #endif
4241
4242 int icmpreplytype4[ICMP_MAXTYPE + 1];
4243
4244
4245 /* ------------------------------------------------------------------------ */
4246 /* Function: fr_matchicmpqueryreply */
4247 /* Returns: int - 1 if "icmp" is a valid reply to "ic" else 0. */
4248 /* Parameters: v(I) - IP protocol version (4 or 6) */
4249 /* ic(I) - ICMP information */
4250 /* icmp(I) - ICMP packet header */
4251 /* rev(I) - direction (0 = forward/1 = reverse) of packet */
4252 /* */
4253 /* Check if the ICMP packet defined by the header pointed to by icmp is a */
4254 /* reply to one as described by what's in ic. If it is a match, return 1, */
4255 /* else return 0 for no match. */
4256 /* ------------------------------------------------------------------------ */
fr_matchicmpqueryreply(v,ic,icmp,rev)4257 int fr_matchicmpqueryreply(v, ic, icmp, rev)
4258 int v;
4259 icmpinfo_t *ic;
4260 icmphdr_t *icmp;
4261 int rev;
4262 {
4263 int ictype;
4264
4265 ictype = ic->ici_type;
4266
4267 if (v == 4) {
4268 /*
4269 * If we matched its type on the way in, then when going out
4270 * it will still be the same type.
4271 */
4272 if ((!rev && (icmp->icmp_type == ictype)) ||
4273 (rev && (icmpreplytype4[ictype] == icmp->icmp_type))) {
4274 if (icmp->icmp_type != ICMP_ECHOREPLY)
4275 return 1;
4276 if (icmp->icmp_id == ic->ici_id)
4277 return 1;
4278 }
4279 }
4280 #ifdef USE_INET6
4281 else if (v == 6) {
4282 if ((!rev && (icmp->icmp_type == ictype)) ||
4283 (rev && (icmpreplytype6[ictype] == icmp->icmp_type))) {
4284 if (icmp->icmp_type != ICMP6_ECHO_REPLY)
4285 return 1;
4286 if (icmp->icmp_id == ic->ici_id)
4287 return 1;
4288 }
4289 }
4290 #endif
4291 return 0;
4292 }
4293
4294
4295 #ifdef IPFILTER_LOOKUP
4296 /* ------------------------------------------------------------------------ */
4297 /* Function: fr_resolvelookup */
4298 /* Returns: void * - NULL = failure, else success. */
4299 /* Parameters: type(I) - type of lookup these parameters are for. */
4300 /* number(I) - table number to use when searching */
4301 /* funcptr(IO) - pointer to pointer for storing IP address */
4302 /* searching function. */
4303 /* ifs - ipf stack instance */
4304 /* */
4305 /* Search for the "table" number passed in amongst those configured for */
4306 /* that particular type. If the type is recognised then the function to */
4307 /* call to do the IP address search will be change, regardless of whether */
4308 /* or not the "table" number exists. */
4309 /* ------------------------------------------------------------------------ */
fr_resolvelookup(type,number,funcptr,ifs)4310 static void *fr_resolvelookup(type, number, funcptr, ifs)
4311 u_int type, number;
4312 lookupfunc_t *funcptr;
4313 ipf_stack_t *ifs;
4314 {
4315 char name[FR_GROUPLEN];
4316 iphtable_t *iph;
4317 ip_pool_t *ipo;
4318 void *ptr;
4319
4320 #if defined(SNPRINTF) && defined(_KERNEL)
4321 (void) SNPRINTF(name, sizeof(name), "%u", number);
4322 #else
4323 (void) sprintf(name, "%u", number);
4324 #endif
4325
4326 READ_ENTER(&ifs->ifs_ip_poolrw);
4327
4328 switch (type)
4329 {
4330 case IPLT_POOL :
4331 # if (defined(__osf__) && defined(_KERNEL))
4332 ptr = NULL;
4333 *funcptr = NULL;
4334 # else
4335 ipo = ip_pool_find(IPL_LOGIPF, name, ifs);
4336 ptr = ipo;
4337 if (ipo != NULL) {
4338 ATOMIC_INC32(ipo->ipo_ref);
4339 }
4340 *funcptr = ip_pool_search;
4341 # endif
4342 break;
4343 case IPLT_HASH :
4344 iph = fr_findhtable(IPL_LOGIPF, name, ifs);
4345 ptr = iph;
4346 if (iph != NULL) {
4347 ATOMIC_INC32(iph->iph_ref);
4348 }
4349 *funcptr = fr_iphmfindip;
4350 break;
4351 default:
4352 ptr = NULL;
4353 *funcptr = NULL;
4354 break;
4355 }
4356 RWLOCK_EXIT(&ifs->ifs_ip_poolrw);
4357
4358 return ptr;
4359 }
4360 #endif
4361
4362
4363 /* ------------------------------------------------------------------------ */
4364 /* Function: frrequest */
4365 /* Returns: int - 0 == success, > 0 == errno value */
4366 /* Parameters: unit(I) - device for which this is for */
4367 /* req(I) - ioctl command (SIOC*) */
4368 /* data(I) - pointr to ioctl data */
4369 /* set(I) - 1 or 0 (filter set) */
4370 /* makecopy(I) - flag indicating whether data points to a rule */
4371 /* in kernel space & hence doesn't need copying. */
4372 /* */
4373 /* This function handles all the requests which operate on the list of */
4374 /* filter rules. This includes adding, deleting, insertion. It is also */
4375 /* responsible for creating groups when a "head" rule is loaded. Interface */
4376 /* names are resolved here and other sanity checks are made on the content */
4377 /* of the rule structure being loaded. If a rule has user defined timeouts */
4378 /* then make sure they are created and initialised before exiting. */
4379 /* ------------------------------------------------------------------------ */
frrequest(unit,req,data,set,makecopy,ifs)4380 int frrequest(unit, req, data, set, makecopy, ifs)
4381 int unit;
4382 ioctlcmd_t req;
4383 int set, makecopy;
4384 caddr_t data;
4385 ipf_stack_t *ifs;
4386 {
4387 frentry_t frd, *fp, *f, **fprev, **ftail;
4388 int error = 0, in, v;
4389 void *ptr, *uptr;
4390 u_int *p, *pp;
4391 frgroup_t *fg;
4392 char *group;
4393
4394 fg = NULL;
4395 fp = &frd;
4396 if (makecopy != 0) {
4397 error = fr_inobj(data, fp, IPFOBJ_FRENTRY);
4398 if (error)
4399 return EFAULT;
4400 if ((fp->fr_flags & FR_T_BUILTIN) != 0)
4401 return EINVAL;
4402 fp->fr_ref = 0;
4403 fp->fr_flags |= FR_COPIED;
4404 } else {
4405 fp = (frentry_t *)data;
4406 if ((fp->fr_type & FR_T_BUILTIN) == 0)
4407 return EINVAL;
4408 fp->fr_flags &= ~FR_COPIED;
4409 }
4410
4411 if (((fp->fr_dsize == 0) && (fp->fr_data != NULL)) ||
4412 ((fp->fr_dsize != 0) && (fp->fr_data == NULL)))
4413 return EINVAL;
4414
4415 v = fp->fr_v;
4416 uptr = fp->fr_data;
4417
4418 /*
4419 * Only filter rules for IPv4 or IPv6 are accepted.
4420 */
4421 if (v == 4)
4422 /*EMPTY*/;
4423 #ifdef USE_INET6
4424 else if (v == 6)
4425 /*EMPTY*/;
4426 #endif
4427 else {
4428 return EINVAL;
4429 }
4430
4431 /*
4432 * If the rule is being loaded from user space, i.e. we had to copy it
4433 * into kernel space, then do not trust the function pointer in the
4434 * rule.
4435 */
4436 if ((makecopy == 1) && (fp->fr_func != NULL)) {
4437 if (fr_findfunc(fp->fr_func) == NULL)
4438 return ESRCH;
4439 error = fr_funcinit(fp, ifs);
4440 if (error != 0)
4441 return error;
4442 }
4443
4444 ptr = NULL;
4445 /*
4446 * Check that the group number does exist and that its use (in/out)
4447 * matches what the rule is.
4448 */
4449 if (!strncmp(fp->fr_grhead, "0", FR_GROUPLEN))
4450 *fp->fr_grhead = '\0';
4451 group = fp->fr_group;
4452 if (!strncmp(group, "0", FR_GROUPLEN))
4453 *group = '\0';
4454
4455 if (FR_ISACCOUNT(fp->fr_flags))
4456 unit = IPL_LOGCOUNT;
4457
4458 if ((req != (int)SIOCZRLST) && (*group != '\0')) {
4459 fg = fr_findgroup(group, unit, set, NULL, ifs);
4460 if (fg == NULL)
4461 return ESRCH;
4462 if (fg->fg_flags == 0)
4463 fg->fg_flags = fp->fr_flags & FR_INOUT;
4464 else if (fg->fg_flags != (fp->fr_flags & FR_INOUT))
4465 return ESRCH;
4466 }
4467
4468 in = (fp->fr_flags & FR_INQUE) ? 0 : 1;
4469
4470 /*
4471 * Work out which rule list this change is being applied to.
4472 */
4473 ftail = NULL;
4474 fprev = NULL;
4475 if (unit == IPL_LOGAUTH)
4476 fprev = &ifs->ifs_ipauth;
4477 else if (v == 4) {
4478 if (FR_ISACCOUNT(fp->fr_flags))
4479 fprev = &ifs->ifs_ipacct[in][set];
4480 else if ((fp->fr_flags & (FR_OUTQUE|FR_INQUE)) != 0)
4481 fprev = &ifs->ifs_ipfilter[in][set];
4482 } else if (v == 6) {
4483 if (FR_ISACCOUNT(fp->fr_flags))
4484 fprev = &ifs->ifs_ipacct6[in][set];
4485 else if ((fp->fr_flags & (FR_OUTQUE|FR_INQUE)) != 0)
4486 fprev = &ifs->ifs_ipfilter6[in][set];
4487 }
4488 if (fprev == NULL)
4489 return ESRCH;
4490
4491 if (*group != '\0') {
4492 if (!fg && !(fg = fr_findgroup(group, unit, set, NULL, ifs)))
4493 return ESRCH;
4494 fprev = &fg->fg_start;
4495 }
4496
4497 ftail = fprev;
4498 for (f = *ftail; (f = *ftail) != NULL; ftail = &f->fr_next) {
4499 if (fp->fr_collect <= f->fr_collect) {
4500 ftail = fprev;
4501 f = NULL;
4502 break;
4503 }
4504 fprev = ftail;
4505 }
4506
4507 /*
4508 * Copy in extra data for the rule.
4509 */
4510 if (fp->fr_dsize != 0) {
4511 if (makecopy != 0) {
4512 KMALLOCS(ptr, void *, fp->fr_dsize);
4513 if (!ptr)
4514 return ENOMEM;
4515 error = COPYIN(uptr, ptr, fp->fr_dsize);
4516 } else {
4517 ptr = uptr;
4518 error = 0;
4519 }
4520 if (error != 0) {
4521 KFREES(ptr, fp->fr_dsize);
4522 return EFAULT;
4523 }
4524 fp->fr_data = ptr;
4525 } else
4526 fp->fr_data = NULL;
4527
4528 /*
4529 * Perform per-rule type sanity checks of their members.
4530 */
4531 switch (fp->fr_type & ~FR_T_BUILTIN)
4532 {
4533 #if defined(IPFILTER_BPF)
4534 case FR_T_BPFOPC :
4535 if (fp->fr_dsize == 0)
4536 return EINVAL;
4537 if (!bpf_validate(ptr, fp->fr_dsize/sizeof(struct bpf_insn))) {
4538 if (makecopy && fp->fr_data != NULL) {
4539 KFREES(fp->fr_data, fp->fr_dsize);
4540 }
4541 return EINVAL;
4542 }
4543 break;
4544 #endif
4545 case FR_T_IPF :
4546 if (fp->fr_dsize != sizeof(fripf_t)) {
4547 if (makecopy && fp->fr_data != NULL) {
4548 KFREES(fp->fr_data, fp->fr_dsize);
4549 }
4550 return EINVAL;
4551 }
4552
4553 /*
4554 * Allowing a rule with both "keep state" and "with oow" is
4555 * pointless because adding a state entry to the table will
4556 * fail with the out of window (oow) flag set.
4557 */
4558 if ((fp->fr_flags & FR_KEEPSTATE) && (fp->fr_flx & FI_OOW)) {
4559 if (makecopy && fp->fr_data != NULL) {
4560 KFREES(fp->fr_data, fp->fr_dsize);
4561 }
4562 return EINVAL;
4563 }
4564
4565 switch (fp->fr_satype)
4566 {
4567 case FRI_BROADCAST :
4568 case FRI_DYNAMIC :
4569 case FRI_NETWORK :
4570 case FRI_NETMASKED :
4571 case FRI_PEERADDR :
4572 if (fp->fr_sifpidx < 0 || fp->fr_sifpidx > 3) {
4573 if (makecopy && fp->fr_data != NULL) {
4574 KFREES(fp->fr_data, fp->fr_dsize);
4575 }
4576 return EINVAL;
4577 }
4578 break;
4579 #ifdef IPFILTER_LOOKUP
4580 case FRI_LOOKUP :
4581 fp->fr_srcptr = fr_resolvelookup(fp->fr_srctype,
4582 fp->fr_srcnum,
4583 &fp->fr_srcfunc, ifs);
4584 break;
4585 #endif
4586 default :
4587 break;
4588 }
4589
4590 switch (fp->fr_datype)
4591 {
4592 case FRI_BROADCAST :
4593 case FRI_DYNAMIC :
4594 case FRI_NETWORK :
4595 case FRI_NETMASKED :
4596 case FRI_PEERADDR :
4597 if (fp->fr_difpidx < 0 || fp->fr_difpidx > 3) {
4598 if (makecopy && fp->fr_data != NULL) {
4599 KFREES(fp->fr_data, fp->fr_dsize);
4600 }
4601 return EINVAL;
4602 }
4603 break;
4604 #ifdef IPFILTER_LOOKUP
4605 case FRI_LOOKUP :
4606 fp->fr_dstptr = fr_resolvelookup(fp->fr_dsttype,
4607 fp->fr_dstnum,
4608 &fp->fr_dstfunc, ifs);
4609 break;
4610 #endif
4611 default :
4612 break;
4613 }
4614 break;
4615 case FR_T_NONE :
4616 break;
4617 case FR_T_CALLFUNC :
4618 break;
4619 case FR_T_COMPIPF :
4620 break;
4621 default :
4622 if (makecopy && fp->fr_data != NULL) {
4623 KFREES(fp->fr_data, fp->fr_dsize);
4624 }
4625 return EINVAL;
4626 }
4627
4628 /*
4629 * Lookup all the interface names that are part of the rule.
4630 */
4631 frsynclist(0, 0, NULL, NULL, fp, ifs);
4632 fp->fr_statecnt = 0;
4633
4634 /*
4635 * Look for an existing matching filter rule, but don't include the
4636 * next or interface pointer in the comparison (fr_next, fr_ifa).
4637 * This elminates rules which are indentical being loaded. Checksum
4638 * the constant part of the filter rule to make comparisons quicker
4639 * (this meaning no pointers are included).
4640 */
4641 for (fp->fr_cksum = 0, p = (u_int *)&fp->fr_func, pp = &fp->fr_cksum;
4642 p < pp; p++)
4643 fp->fr_cksum += *p;
4644 pp = (u_int *)(fp->fr_caddr + fp->fr_dsize);
4645 for (p = (u_int *)fp->fr_data; p < pp; p++)
4646 fp->fr_cksum += *p;
4647
4648 WRITE_ENTER(&ifs->ifs_ipf_mutex);
4649 bzero((char *)ifs->ifs_frcache, sizeof (ifs->ifs_frcache));
4650
4651 for (; (f = *ftail) != NULL; ftail = &f->fr_next) {
4652 if ((fp->fr_cksum != f->fr_cksum) ||
4653 (f->fr_dsize != fp->fr_dsize))
4654 continue;
4655 if (bcmp((char *)&f->fr_func, (char *)&fp->fr_func, FR_CMPSIZ))
4656 continue;
4657 if ((!ptr && !f->fr_data) ||
4658 (ptr && f->fr_data &&
4659 !bcmp((char *)ptr, (char *)f->fr_data, f->fr_dsize)))
4660 break;
4661 }
4662
4663 /*
4664 * If zero'ing statistics, copy current to caller and zero.
4665 */
4666 if (req == (ioctlcmd_t)SIOCZRLST) {
4667 if (f == NULL)
4668 error = ESRCH;
4669 else {
4670 /*
4671 * Copy and reduce lock because of impending copyout.
4672 * Well we should, but if we do then the atomicity of
4673 * this call and the correctness of fr_hits and
4674 * fr_bytes cannot be guaranteed. As it is, this code
4675 * only resets them to 0 if they are successfully
4676 * copied out into user space.
4677 */
4678 bcopy((char *)f, (char *)fp, sizeof(*f));
4679
4680 /*
4681 * When we copy this rule back out, set the data
4682 * pointer to be what it was in user space.
4683 */
4684 fp->fr_data = uptr;
4685 error = fr_outobj(data, fp, IPFOBJ_FRENTRY);
4686
4687 if (error == 0) {
4688 if ((f->fr_dsize != 0) && (uptr != NULL))
4689 error = COPYOUT(f->fr_data, uptr,
4690 f->fr_dsize);
4691 if (error == 0) {
4692 f->fr_hits = 0;
4693 f->fr_bytes = 0;
4694 }
4695 }
4696 }
4697
4698 if ((ptr != NULL) && (makecopy != 0)) {
4699 KFREES(ptr, fp->fr_dsize);
4700 }
4701 RWLOCK_EXIT(&ifs->ifs_ipf_mutex);
4702 return error;
4703 }
4704
4705 if (!f) {
4706 /*
4707 * At the end of this, ftail must point to the place where the
4708 * new rule is to be saved/inserted/added.
4709 * For SIOCAD*FR, this should be the last rule in the group of
4710 * rules that have equal fr_collect fields.
4711 * For SIOCIN*FR, ...
4712 */
4713 if (req == (ioctlcmd_t)SIOCADAFR ||
4714 req == (ioctlcmd_t)SIOCADIFR) {
4715
4716 for (ftail = fprev; (f = *ftail) != NULL; ) {
4717 if (f->fr_collect > fp->fr_collect)
4718 break;
4719 ftail = &f->fr_next;
4720 }
4721 f = NULL;
4722 ptr = NULL;
4723 error = 0;
4724 } else if (req == (ioctlcmd_t)SIOCINAFR ||
4725 req == (ioctlcmd_t)SIOCINIFR) {
4726 while ((f = *fprev) != NULL) {
4727 if (f->fr_collect >= fp->fr_collect)
4728 break;
4729 fprev = &f->fr_next;
4730 }
4731 ftail = fprev;
4732 if (fp->fr_hits != 0) {
4733 while (fp->fr_hits && (f = *ftail)) {
4734 if (f->fr_collect != fp->fr_collect)
4735 break;
4736 fprev = ftail;
4737 ftail = &f->fr_next;
4738 fp->fr_hits--;
4739 }
4740 }
4741 f = NULL;
4742 ptr = NULL;
4743 error = 0;
4744 }
4745 }
4746
4747 /*
4748 * Request to remove a rule.
4749 */
4750 if (req == (ioctlcmd_t)SIOCRMAFR || req == (ioctlcmd_t)SIOCRMIFR) {
4751 if (!f)
4752 error = ESRCH;
4753 else {
4754 /*
4755 * Do not allow activity from user space to interfere
4756 * with rules not loaded that way.
4757 */
4758 if ((makecopy == 1) && !(f->fr_flags & FR_COPIED)) {
4759 error = EPERM;
4760 goto done;
4761 }
4762
4763 /*
4764 * Return EBUSY if the rule is being reference by
4765 * something else (eg state information.
4766 */
4767 if (f->fr_ref > 1) {
4768 error = EBUSY;
4769 goto done;
4770 }
4771 #ifdef IPFILTER_SCAN
4772 if (f->fr_isctag[0] != '\0' &&
4773 (f->fr_isc != (struct ipscan *)-1))
4774 ipsc_detachfr(f);
4775 #endif
4776 if (unit == IPL_LOGAUTH) {
4777 error = fr_preauthcmd(req, f, ftail, ifs);
4778 goto done;
4779 }
4780 if (*f->fr_grhead != '\0')
4781 fr_delgroup(f->fr_grhead, unit, set, ifs);
4782 fr_fixskip(ftail, f, -1);
4783 *ftail = f->fr_next;
4784 f->fr_next = NULL;
4785 (void)fr_derefrule(&f, ifs);
4786 }
4787 } else {
4788 /*
4789 * Not removing, so we must be adding/inserting a rule.
4790 */
4791 if (f)
4792 error = EEXIST;
4793 else {
4794 if (unit == IPL_LOGAUTH) {
4795 error = fr_preauthcmd(req, fp, ftail, ifs);
4796 goto done;
4797 }
4798 if (makecopy) {
4799 KMALLOC(f, frentry_t *);
4800 } else
4801 f = fp;
4802 if (f != NULL) {
4803 if (fp != f)
4804 bcopy((char *)fp, (char *)f,
4805 sizeof(*f));
4806 MUTEX_NUKE(&f->fr_lock);
4807 MUTEX_INIT(&f->fr_lock, "filter rule lock");
4808 #ifdef IPFILTER_SCAN
4809 if (f->fr_isctag[0] != '\0' &&
4810 ipsc_attachfr(f))
4811 f->fr_isc = (struct ipscan *)-1;
4812 #endif
4813 f->fr_hits = 0;
4814 if (makecopy != 0)
4815 f->fr_ref = 1;
4816 f->fr_next = *ftail;
4817 *ftail = f;
4818 if (req == (ioctlcmd_t)SIOCINIFR ||
4819 req == (ioctlcmd_t)SIOCINAFR)
4820 fr_fixskip(ftail, f, 1);
4821 f->fr_grp = NULL;
4822 group = f->fr_grhead;
4823 if (*group != '\0') {
4824 fg = fr_addgroup(group, f, f->fr_flags,
4825 unit, set, ifs);
4826 if (fg != NULL)
4827 f->fr_grp = &fg->fg_start;
4828 }
4829 } else
4830 error = ENOMEM;
4831 }
4832 }
4833 done:
4834 RWLOCK_EXIT(&ifs->ifs_ipf_mutex);
4835 if ((ptr != NULL) && (error != 0) && (makecopy != 0)) {
4836 KFREES(ptr, fp->fr_dsize);
4837 }
4838 return (error);
4839 }
4840
4841
4842 /* ------------------------------------------------------------------------ */
4843 /* Function: fr_funcinit */
4844 /* Returns: int - 0 == success, else ESRCH: cannot resolve rule details */
4845 /* Parameters: fr(I) - pointer to filter rule */
4846 /* */
4847 /* If a rule is a call rule, then check if the function it points to needs */
4848 /* an init function to be called now the rule has been loaded. */
4849 /* ------------------------------------------------------------------------ */
fr_funcinit(fr,ifs)4850 static int fr_funcinit(fr, ifs)
4851 frentry_t *fr;
4852 ipf_stack_t *ifs;
4853 {
4854 ipfunc_resolve_t *ft;
4855 int err;
4856
4857 err = ESRCH;
4858
4859 for (ft = fr_availfuncs; ft->ipfu_addr != NULL; ft++)
4860 if (ft->ipfu_addr == fr->fr_func) {
4861 err = 0;
4862 if (ft->ipfu_init != NULL)
4863 err = (*ft->ipfu_init)(fr, ifs);
4864 break;
4865 }
4866 return err;
4867 }
4868
4869
4870 /* ------------------------------------------------------------------------ */
4871 /* Function: fr_findfunc */
4872 /* Returns: ipfunc_t - pointer to function if found, else NULL */
4873 /* Parameters: funcptr(I) - function pointer to lookup */
4874 /* */
4875 /* Look for a function in the table of known functions. */
4876 /* ------------------------------------------------------------------------ */
fr_findfunc(funcptr)4877 static ipfunc_t fr_findfunc(funcptr)
4878 ipfunc_t funcptr;
4879 {
4880 ipfunc_resolve_t *ft;
4881
4882 for (ft = fr_availfuncs; ft->ipfu_addr != NULL; ft++)
4883 if (ft->ipfu_addr == funcptr)
4884 return funcptr;
4885 return NULL;
4886 }
4887
4888
4889 /* ------------------------------------------------------------------------ */
4890 /* Function: fr_resolvefunc */
4891 /* Returns: int - 0 == success, else error */
4892 /* Parameters: data(IO) - ioctl data pointer to ipfunc_resolve_t struct */
4893 /* */
4894 /* Copy in a ipfunc_resolve_t structure and then fill in the missing field. */
4895 /* This will either be the function name (if the pointer is set) or the */
4896 /* function pointer if the name is set. When found, fill in the other one */
4897 /* so that the entire, complete, structure can be copied back to user space.*/
4898 /* ------------------------------------------------------------------------ */
fr_resolvefunc(data)4899 int fr_resolvefunc(data)
4900 void *data;
4901 {
4902 ipfunc_resolve_t res, *ft;
4903 int err;
4904
4905 err = BCOPYIN(data, &res, sizeof(res));
4906 if (err != 0)
4907 return EFAULT;
4908
4909 if (res.ipfu_addr == NULL && res.ipfu_name[0] != '\0') {
4910 for (ft = fr_availfuncs; ft->ipfu_addr != NULL; ft++)
4911 if (strncmp(res.ipfu_name, ft->ipfu_name,
4912 sizeof(res.ipfu_name)) == 0) {
4913 res.ipfu_addr = ft->ipfu_addr;
4914 res.ipfu_init = ft->ipfu_init;
4915 if (COPYOUT(&res, data, sizeof(res)) != 0)
4916 return EFAULT;
4917 return 0;
4918 }
4919 }
4920 if (res.ipfu_addr != NULL && res.ipfu_name[0] == '\0') {
4921 for (ft = fr_availfuncs; ft->ipfu_addr != NULL; ft++)
4922 if (ft->ipfu_addr == res.ipfu_addr) {
4923 (void) strncpy(res.ipfu_name, ft->ipfu_name,
4924 sizeof(res.ipfu_name));
4925 res.ipfu_init = ft->ipfu_init;
4926 if (COPYOUT(&res, data, sizeof(res)) != 0)
4927 return EFAULT;
4928 return 0;
4929 }
4930 }
4931 return ESRCH;
4932 }
4933
4934
4935 #if !defined(_KERNEL) || (!defined(__NetBSD__) && !defined(__OpenBSD__) && !defined(__FreeBSD__)) || \
4936 (defined(__FreeBSD__) && (__FreeBSD_version < 490000)) || \
4937 (defined(__NetBSD__) && (__NetBSD_Version__ < 105000000)) || \
4938 (defined(__OpenBSD__) && (OpenBSD < 200006))
4939 /*
4940 * From: NetBSD
4941 * ppsratecheck(): packets (or events) per second limitation.
4942 */
4943 int
ppsratecheck(lasttime,curpps,maxpps)4944 ppsratecheck(lasttime, curpps, maxpps)
4945 struct timeval *lasttime;
4946 int *curpps;
4947 int maxpps; /* maximum pps allowed */
4948 {
4949 struct timeval tv, delta;
4950 int rv;
4951
4952 GETKTIME(&tv);
4953
4954 delta.tv_sec = tv.tv_sec - lasttime->tv_sec;
4955 delta.tv_usec = tv.tv_usec - lasttime->tv_usec;
4956 if (delta.tv_usec < 0) {
4957 delta.tv_sec--;
4958 delta.tv_usec += 1000000;
4959 }
4960
4961 /*
4962 * check for 0,0 is so that the message will be seen at least once.
4963 * if more than one second have passed since the last update of
4964 * lasttime, reset the counter.
4965 *
4966 * we do increment *curpps even in *curpps < maxpps case, as some may
4967 * try to use *curpps for stat purposes as well.
4968 */
4969 if ((lasttime->tv_sec == 0 && lasttime->tv_usec == 0) ||
4970 delta.tv_sec >= 1) {
4971 *lasttime = tv;
4972 *curpps = 0;
4973 rv = 1;
4974 } else if (maxpps < 0)
4975 rv = 1;
4976 else if (*curpps < maxpps)
4977 rv = 1;
4978 else
4979 rv = 0;
4980 *curpps = *curpps + 1;
4981
4982 return (rv);
4983 }
4984 #endif
4985
4986
4987 /* ------------------------------------------------------------------------ */
4988 /* Function: fr_derefrule */
4989 /* Returns: int - 0 == rule freed up, else rule not freed */
4990 /* Parameters: fr(I) - pointer to filter rule */
4991 /* */
4992 /* Decrement the reference counter to a rule by one. If it reaches zero, */
4993 /* free it and any associated storage space being used by it. */
4994 /* ------------------------------------------------------------------------ */
fr_derefrule(frp,ifs)4995 int fr_derefrule(frp, ifs)
4996 frentry_t **frp;
4997 ipf_stack_t *ifs;
4998 {
4999 frentry_t *fr;
5000
5001 fr = *frp;
5002
5003 MUTEX_ENTER(&fr->fr_lock);
5004 fr->fr_ref--;
5005 if (fr->fr_ref == 0) {
5006 MUTEX_EXIT(&fr->fr_lock);
5007 MUTEX_DESTROY(&fr->fr_lock);
5008
5009 #ifdef IPFILTER_LOOKUP
5010 if (fr->fr_type == FR_T_IPF && fr->fr_satype == FRI_LOOKUP)
5011 ip_lookup_deref(fr->fr_srctype, fr->fr_srcptr, ifs);
5012 if (fr->fr_type == FR_T_IPF && fr->fr_datype == FRI_LOOKUP)
5013 ip_lookup_deref(fr->fr_dsttype, fr->fr_dstptr, ifs);
5014 #endif
5015
5016 if (fr->fr_dsize) {
5017 KFREES(fr->fr_data, fr->fr_dsize);
5018 }
5019 if ((fr->fr_flags & FR_COPIED) != 0) {
5020 KFREE(fr);
5021 return 0;
5022 }
5023 return 1;
5024 } else {
5025 MUTEX_EXIT(&fr->fr_lock);
5026 }
5027 *frp = NULL;
5028 return -1;
5029 }
5030
5031
5032 #ifdef IPFILTER_LOOKUP
5033 /* ------------------------------------------------------------------------ */
5034 /* Function: fr_grpmapinit */
5035 /* Returns: int - 0 == success, else ESRCH because table entry not found*/
5036 /* Parameters: fr(I) - pointer to rule to find hash table for */
5037 /* */
5038 /* Looks for group hash table fr_arg and stores a pointer to it in fr_ptr. */
5039 /* fr_ptr is later used by fr_srcgrpmap and fr_dstgrpmap. */
5040 /* ------------------------------------------------------------------------ */
fr_grpmapinit(fr,ifs)5041 static int fr_grpmapinit(fr, ifs)
5042 frentry_t *fr;
5043 ipf_stack_t *ifs;
5044 {
5045 char name[FR_GROUPLEN];
5046 iphtable_t *iph;
5047
5048 #if defined(SNPRINTF) && defined(_KERNEL)
5049 (void) SNPRINTF(name, sizeof(name), "%d", fr->fr_arg);
5050 #else
5051 (void) sprintf(name, "%d", fr->fr_arg);
5052 #endif
5053 iph = fr_findhtable(IPL_LOGIPF, name, ifs);
5054 if (iph == NULL)
5055 return ESRCH;
5056 if ((iph->iph_flags & FR_INOUT) != (fr->fr_flags & FR_INOUT))
5057 return ESRCH;
5058 fr->fr_ptr = iph;
5059 return 0;
5060 }
5061
5062
5063 /* ------------------------------------------------------------------------ */
5064 /* Function: fr_srcgrpmap */
5065 /* Returns: frentry_t * - pointer to "new last matching" rule or NULL */
5066 /* Parameters: fin(I) - pointer to packet information */
5067 /* passp(IO) - pointer to current/new filter decision (unused) */
5068 /* */
5069 /* Look for a rule group head in a hash table, using the source address as */
5070 /* the key, and descend into that group and continue matching rules against */
5071 /* the packet. */
5072 /* ------------------------------------------------------------------------ */
fr_srcgrpmap(fin,passp)5073 frentry_t *fr_srcgrpmap(fin, passp)
5074 fr_info_t *fin;
5075 u_32_t *passp;
5076 {
5077 frgroup_t *fg;
5078 void *rval;
5079 ipf_stack_t *ifs = fin->fin_ifs;
5080
5081 rval = fr_iphmfindgroup(fin->fin_fr->fr_ptr, fin->fin_v, &fin->fin_src, ifs);
5082 if (rval == NULL)
5083 return NULL;
5084
5085 fg = rval;
5086 fin->fin_fr = fg->fg_start;
5087 (void) fr_scanlist(fin, *passp);
5088 return fin->fin_fr;
5089 }
5090
5091
5092 /* ------------------------------------------------------------------------ */
5093 /* Function: fr_dstgrpmap */
5094 /* Returns: frentry_t * - pointer to "new last matching" rule or NULL */
5095 /* Parameters: fin(I) - pointer to packet information */
5096 /* passp(IO) - pointer to current/new filter decision (unused) */
5097 /* */
5098 /* Look for a rule group head in a hash table, using the destination */
5099 /* address as the key, and descend into that group and continue matching */
5100 /* rules against the packet. */
5101 /* ------------------------------------------------------------------------ */
fr_dstgrpmap(fin,passp)5102 frentry_t *fr_dstgrpmap(fin, passp)
5103 fr_info_t *fin;
5104 u_32_t *passp;
5105 {
5106 frgroup_t *fg;
5107 void *rval;
5108 ipf_stack_t *ifs = fin->fin_ifs;
5109
5110 rval = fr_iphmfindgroup(fin->fin_fr->fr_ptr, fin->fin_v, &fin->fin_dst, ifs);
5111 if (rval == NULL)
5112 return NULL;
5113
5114 fg = rval;
5115 fin->fin_fr = fg->fg_start;
5116 (void) fr_scanlist(fin, *passp);
5117 return fin->fin_fr;
5118 }
5119 #endif /* IPFILTER_LOOKUP */
5120
5121 /*
5122 * Queue functions
5123 * ===============
5124 * These functions manage objects on queues for efficient timeouts. There are
5125 * a number of system defined queues as well as user defined timeouts. It is
5126 * expected that a lock is held in the domain in which the queue belongs
5127 * (i.e. either state or NAT) when calling any of these functions that prevents
5128 * fr_freetimeoutqueue() from being called at the same time as any other.
5129 */
5130
5131
5132 /* ------------------------------------------------------------------------ */
5133 /* Function: fr_addtimeoutqueue */
5134 /* Returns: struct ifqtq * - NULL if malloc fails, else pointer to */
5135 /* timeout queue with given interval. */
5136 /* Parameters: parent(I) - pointer to pointer to parent node of this list */
5137 /* of interface queues. */
5138 /* seconds(I) - timeout value in seconds for this queue. */
5139 /* */
5140 /* This routine first looks for a timeout queue that matches the interval */
5141 /* being requested. If it finds one, increments the reference counter and */
5142 /* returns a pointer to it. If none are found, it allocates a new one and */
5143 /* inserts it at the top of the list. */
5144 /* */
5145 /* Locking. */
5146 /* It is assumed that the caller of this function has an appropriate lock */
5147 /* held (exclusively) in the domain that encompases 'parent'. */
5148 /* ------------------------------------------------------------------------ */
fr_addtimeoutqueue(parent,seconds,ifs)5149 ipftq_t *fr_addtimeoutqueue(parent, seconds, ifs)
5150 ipftq_t **parent;
5151 u_int seconds;
5152 ipf_stack_t *ifs;
5153 {
5154 ipftq_t *ifq;
5155 u_int period;
5156
5157 period = seconds * IPF_HZ_DIVIDE;
5158
5159 MUTEX_ENTER(&ifs->ifs_ipf_timeoutlock);
5160 for (ifq = *parent; ifq != NULL; ifq = ifq->ifq_next) {
5161 if (ifq->ifq_ttl == period) {
5162 /*
5163 * Reset the delete flag, if set, so the structure
5164 * gets reused rather than freed and reallocated.
5165 */
5166 MUTEX_ENTER(&ifq->ifq_lock);
5167 ifq->ifq_flags &= ~IFQF_DELETE;
5168 ifq->ifq_ref++;
5169 MUTEX_EXIT(&ifq->ifq_lock);
5170 MUTEX_EXIT(&ifs->ifs_ipf_timeoutlock);
5171
5172 return ifq;
5173 }
5174 }
5175
5176 KMALLOC(ifq, ipftq_t *);
5177 if (ifq != NULL) {
5178 ifq->ifq_ttl = period;
5179 ifq->ifq_head = NULL;
5180 ifq->ifq_tail = &ifq->ifq_head;
5181 ifq->ifq_next = *parent;
5182 ifq->ifq_pnext = parent;
5183 ifq->ifq_ref = 1;
5184 ifq->ifq_flags = IFQF_USER;
5185 *parent = ifq;
5186 ifs->ifs_fr_userifqs++;
5187 MUTEX_NUKE(&ifq->ifq_lock);
5188 MUTEX_INIT(&ifq->ifq_lock, "ipftq mutex");
5189 }
5190 MUTEX_EXIT(&ifs->ifs_ipf_timeoutlock);
5191 return ifq;
5192 }
5193
5194
5195 /* ------------------------------------------------------------------------ */
5196 /* Function: fr_deletetimeoutqueue */
5197 /* Returns: int - new reference count value of the timeout queue */
5198 /* Parameters: ifq(I) - timeout queue which is losing a reference. */
5199 /* Locks: ifq->ifq_lock */
5200 /* */
5201 /* This routine must be called when we're discarding a pointer to a timeout */
5202 /* queue object, taking care of the reference counter. */
5203 /* */
5204 /* Now that this just sets a DELETE flag, it requires the expire code to */
5205 /* check the list of user defined timeout queues and call the free function */
5206 /* below (currently commented out) to stop memory leaking. It is done this */
5207 /* way because the locking may not be sufficient to safely do a free when */
5208 /* this function is called. */
5209 /* ------------------------------------------------------------------------ */
fr_deletetimeoutqueue(ifq)5210 int fr_deletetimeoutqueue(ifq)
5211 ipftq_t *ifq;
5212 {
5213
5214 ifq->ifq_ref--;
5215 if ((ifq->ifq_ref == 0) && ((ifq->ifq_flags & IFQF_USER) != 0)) {
5216 ifq->ifq_flags |= IFQF_DELETE;
5217 }
5218
5219 return ifq->ifq_ref;
5220 }
5221
5222
5223 /* ------------------------------------------------------------------------ */
5224 /* Function: fr_freetimeoutqueue */
5225 /* Parameters: ifq(I) - timeout queue which is losing a reference. */
5226 /* Returns: Nil */
5227 /* */
5228 /* Locking: */
5229 /* It is assumed that the caller of this function has an appropriate lock */
5230 /* held (exclusively) in the domain that encompases the callers "domain". */
5231 /* The ifq_lock for this structure should not be held. */
5232 /* */
5233 /* Remove a user definde timeout queue from the list of queues it is in and */
5234 /* tidy up after this is done. */
5235 /* ------------------------------------------------------------------------ */
fr_freetimeoutqueue(ifq,ifs)5236 void fr_freetimeoutqueue(ifq, ifs)
5237 ipftq_t *ifq;
5238 ipf_stack_t *ifs;
5239 {
5240
5241
5242 if (((ifq->ifq_flags & IFQF_DELETE) == 0) || (ifq->ifq_ref != 0) ||
5243 ((ifq->ifq_flags & IFQF_USER) == 0)) {
5244 printf("fr_freetimeoutqueue(%lx) flags 0x%x ttl %d ref %d\n",
5245 (u_long)ifq, ifq->ifq_flags, ifq->ifq_ttl,
5246 ifq->ifq_ref);
5247 return;
5248 }
5249
5250 /*
5251 * Remove from its position in the list.
5252 */
5253 *ifq->ifq_pnext = ifq->ifq_next;
5254 if (ifq->ifq_next != NULL)
5255 ifq->ifq_next->ifq_pnext = ifq->ifq_pnext;
5256
5257 MUTEX_DESTROY(&ifq->ifq_lock);
5258 ifs->ifs_fr_userifqs--;
5259 KFREE(ifq);
5260 }
5261
5262
5263 /* ------------------------------------------------------------------------ */
5264 /* Function: fr_deletequeueentry */
5265 /* Returns: Nil */
5266 /* Parameters: tqe(I) - timeout queue entry to delete */
5267 /* ifq(I) - timeout queue to remove entry from */
5268 /* */
5269 /* Remove a tail queue entry from its queue and make it an orphan. */
5270 /* fr_deletetimeoutqueue is called to make sure the reference count on the */
5271 /* queue is correct. We can't, however, call fr_freetimeoutqueue because */
5272 /* the correct lock(s) may not be held that would make it safe to do so. */
5273 /* ------------------------------------------------------------------------ */
fr_deletequeueentry(tqe)5274 void fr_deletequeueentry(tqe)
5275 ipftqent_t *tqe;
5276 {
5277 ipftq_t *ifq;
5278
5279 ifq = tqe->tqe_ifq;
5280 if (ifq == NULL)
5281 return;
5282
5283 MUTEX_ENTER(&ifq->ifq_lock);
5284
5285 if (tqe->tqe_pnext != NULL) {
5286 *tqe->tqe_pnext = tqe->tqe_next;
5287 if (tqe->tqe_next != NULL)
5288 tqe->tqe_next->tqe_pnext = tqe->tqe_pnext;
5289 else /* we must be the tail anyway */
5290 ifq->ifq_tail = tqe->tqe_pnext;
5291
5292 tqe->tqe_pnext = NULL;
5293 tqe->tqe_ifq = NULL;
5294 }
5295
5296 (void) fr_deletetimeoutqueue(ifq);
5297
5298 MUTEX_EXIT(&ifq->ifq_lock);
5299 }
5300
5301
5302 /* ------------------------------------------------------------------------ */
5303 /* Function: fr_queuefront */
5304 /* Returns: Nil */
5305 /* Parameters: tqe(I) - pointer to timeout queue entry */
5306 /* */
5307 /* Move a queue entry to the front of the queue, if it isn't already there. */
5308 /* ------------------------------------------------------------------------ */
fr_queuefront(tqe)5309 void fr_queuefront(tqe)
5310 ipftqent_t *tqe;
5311 {
5312 ipftq_t *ifq;
5313
5314 ifq = tqe->tqe_ifq;
5315 if (ifq == NULL)
5316 return;
5317
5318 MUTEX_ENTER(&ifq->ifq_lock);
5319 if (ifq->ifq_head != tqe) {
5320 *tqe->tqe_pnext = tqe->tqe_next;
5321 if (tqe->tqe_next)
5322 tqe->tqe_next->tqe_pnext = tqe->tqe_pnext;
5323 else
5324 ifq->ifq_tail = tqe->tqe_pnext;
5325
5326 tqe->tqe_next = ifq->ifq_head;
5327 ifq->ifq_head->tqe_pnext = &tqe->tqe_next;
5328 ifq->ifq_head = tqe;
5329 tqe->tqe_pnext = &ifq->ifq_head;
5330 }
5331 MUTEX_EXIT(&ifq->ifq_lock);
5332 }
5333
5334
5335 /* ------------------------------------------------------------------------ */
5336 /* Function: fr_queueback */
5337 /* Returns: Nil */
5338 /* Parameters: tqe(I) - pointer to timeout queue entry */
5339 /* */
5340 /* Move a queue entry to the back of the queue, if it isn't already there. */
5341 /* ------------------------------------------------------------------------ */
fr_queueback(tqe,ifs)5342 void fr_queueback(tqe, ifs)
5343 ipftqent_t *tqe;
5344 ipf_stack_t *ifs;
5345 {
5346 ipftq_t *ifq;
5347
5348 ifq = tqe->tqe_ifq;
5349 if (ifq == NULL)
5350 return;
5351 tqe->tqe_die = ifs->ifs_fr_ticks + ifq->ifq_ttl;
5352
5353 MUTEX_ENTER(&ifq->ifq_lock);
5354 if (tqe->tqe_next == NULL) { /* at the end already ? */
5355 MUTEX_EXIT(&ifq->ifq_lock);
5356 return;
5357 }
5358
5359 /*
5360 * Remove from list
5361 */
5362 *tqe->tqe_pnext = tqe->tqe_next;
5363 tqe->tqe_next->tqe_pnext = tqe->tqe_pnext;
5364
5365 /*
5366 * Make it the last entry.
5367 */
5368 tqe->tqe_next = NULL;
5369 tqe->tqe_pnext = ifq->ifq_tail;
5370 *ifq->ifq_tail = tqe;
5371 ifq->ifq_tail = &tqe->tqe_next;
5372 MUTEX_EXIT(&ifq->ifq_lock);
5373 }
5374
5375
5376 /* ------------------------------------------------------------------------ */
5377 /* Function: fr_queueappend */
5378 /* Returns: Nil */
5379 /* Parameters: tqe(I) - pointer to timeout queue entry */
5380 /* ifq(I) - pointer to timeout queue */
5381 /* parent(I) - owing object pointer */
5382 /* */
5383 /* Add a new item to this queue and put it on the very end. */
5384 /* ------------------------------------------------------------------------ */
fr_queueappend(tqe,ifq,parent,ifs)5385 void fr_queueappend(tqe, ifq, parent, ifs)
5386 ipftqent_t *tqe;
5387 ipftq_t *ifq;
5388 void *parent;
5389 ipf_stack_t *ifs;
5390 {
5391
5392 MUTEX_ENTER(&ifq->ifq_lock);
5393 tqe->tqe_parent = parent;
5394 tqe->tqe_pnext = ifq->ifq_tail;
5395 *ifq->ifq_tail = tqe;
5396 ifq->ifq_tail = &tqe->tqe_next;
5397 tqe->tqe_next = NULL;
5398 tqe->tqe_ifq = ifq;
5399 tqe->tqe_die = ifs->ifs_fr_ticks + ifq->ifq_ttl;
5400 ifq->ifq_ref++;
5401 MUTEX_EXIT(&ifq->ifq_lock);
5402 }
5403
5404
5405 /* ------------------------------------------------------------------------ */
5406 /* Function: fr_movequeue */
5407 /* Returns: Nil */
5408 /* Parameters: tq(I) - pointer to timeout queue information */
5409 /* oifp(I) - old timeout queue entry was on */
5410 /* nifp(I) - new timeout queue to put entry on */
5411 /* ifs - ipf stack instance */
5412 /* */
5413 /* Move a queue entry from one timeout queue to another timeout queue. */
5414 /* If it notices that the current entry is already last and does not need */
5415 /* to move queue, the return. */
5416 /* ------------------------------------------------------------------------ */
fr_movequeue(tqe,oifq,nifq,ifs)5417 void fr_movequeue(tqe, oifq, nifq, ifs)
5418 ipftqent_t *tqe;
5419 ipftq_t *oifq, *nifq;
5420 ipf_stack_t *ifs;
5421 {
5422 /*
5423 * If the queue isn't changing, and the clock hasn't ticked
5424 * since the last update, the operation will be a no-op.
5425 */
5426 if (oifq == nifq && tqe->tqe_touched == ifs->ifs_fr_ticks)
5427 return;
5428
5429 /*
5430 * Grab the lock and update the timers.
5431 */
5432 MUTEX_ENTER(&oifq->ifq_lock);
5433 tqe->tqe_touched = ifs->ifs_fr_ticks;
5434 tqe->tqe_die = ifs->ifs_fr_ticks + nifq->ifq_ttl;
5435
5436 /*
5437 * The remainder of the operation can still be a no-op.
5438 *
5439 * If the queue isn't changing, check to see if
5440 * an update would be meaningless.
5441 */
5442 if (oifq == nifq) {
5443 if ((tqe->tqe_next == NULL) ||
5444 (tqe->tqe_next->tqe_die == tqe->tqe_die)) {
5445 MUTEX_EXIT(&oifq->ifq_lock);
5446 return;
5447 }
5448 }
5449
5450 /*
5451 * Remove from the old queue
5452 */
5453 *tqe->tqe_pnext = tqe->tqe_next;
5454 if (tqe->tqe_next)
5455 tqe->tqe_next->tqe_pnext = tqe->tqe_pnext;
5456 else
5457 oifq->ifq_tail = tqe->tqe_pnext;
5458 tqe->tqe_next = NULL;
5459
5460 /*
5461 * If we're moving from one queue to another, release the lock on the
5462 * old queue and get a lock on the new queue. For user defined queues,
5463 * if we're moving off it, call delete in case it can now be freed.
5464 */
5465 if (oifq != nifq) {
5466 tqe->tqe_ifq = NULL;
5467
5468 (void) fr_deletetimeoutqueue(oifq);
5469
5470 MUTEX_EXIT(&oifq->ifq_lock);
5471
5472 MUTEX_ENTER(&nifq->ifq_lock);
5473
5474 tqe->tqe_ifq = nifq;
5475 nifq->ifq_ref++;
5476 }
5477
5478 /*
5479 * Add to the bottom of the new queue
5480 */
5481 tqe->tqe_pnext = nifq->ifq_tail;
5482 *nifq->ifq_tail = tqe;
5483 nifq->ifq_tail = &tqe->tqe_next;
5484 MUTEX_EXIT(&nifq->ifq_lock);
5485 }
5486
5487
5488 /* ------------------------------------------------------------------------ */
5489 /* Function: fr_updateipid */
5490 /* Returns: int - 0 == success, -1 == error (packet should be droppped) */
5491 /* Parameters: fin(I) - pointer to packet information */
5492 /* */
5493 /* When we are doing NAT, change the IP of every packet to represent a */
5494 /* single sequence of packets coming from the host, hiding any host */
5495 /* specific sequencing that might otherwise be revealed. If the packet is */
5496 /* a fragment, then store the 'new' IPid in the fragment cache and look up */
5497 /* the fragment cache for non-leading fragments. If a non-leading fragment */
5498 /* has no match in the cache, return an error. */
5499 /* ------------------------------------------------------------------------ */
fr_updateipid(fin)5500 static INLINE int fr_updateipid(fin)
5501 fr_info_t *fin;
5502 {
5503 u_short id, ido, sums;
5504 u_32_t sumd, sum;
5505 ip_t *ip;
5506
5507 if (fin->fin_off != 0) {
5508 sum = fr_ipid_knownfrag(fin);
5509 if (sum == 0xffffffff)
5510 return -1;
5511 sum &= 0xffff;
5512 id = (u_short)sum;
5513 } else {
5514 id = fr_nextipid(fin);
5515 if (fin->fin_off == 0 && (fin->fin_flx & FI_FRAG) != 0)
5516 (void) fr_ipid_newfrag(fin, (u_32_t)id);
5517 }
5518
5519 ip = fin->fin_ip;
5520 ido = ntohs(ip->ip_id);
5521 if (id == ido)
5522 return 0;
5523 ip->ip_id = htons(id);
5524 CALC_SUMD(ido, id, sumd); /* DESTRUCTIVE MACRO! id,ido change */
5525 sum = (~ntohs(ip->ip_sum)) & 0xffff;
5526 sum += sumd;
5527 sum = (sum >> 16) + (sum & 0xffff);
5528 sum = (sum >> 16) + (sum & 0xffff);
5529 sums = ~(u_short)sum;
5530 ip->ip_sum = htons(sums);
5531 return 0;
5532 }
5533
5534
5535 #ifdef NEED_FRGETIFNAME
5536 /* ------------------------------------------------------------------------ */
5537 /* Function: fr_getifname */
5538 /* Returns: char * - pointer to interface name */
5539 /* Parameters: ifp(I) - pointer to network interface */
5540 /* buffer(O) - pointer to where to store interface name */
5541 /* */
5542 /* Constructs an interface name in the buffer passed. The buffer passed is */
5543 /* expected to be at least LIFNAMSIZ in bytes big. If buffer is passed in */
5544 /* as a NULL pointer then return a pointer to a static array. */
5545 /* ------------------------------------------------------------------------ */
fr_getifname(ifp,buffer)5546 char *fr_getifname(ifp, buffer)
5547 struct ifnet *ifp;
5548 char *buffer;
5549 {
5550 static char namebuf[LIFNAMSIZ];
5551 # if defined(MENTAT) || defined(__FreeBSD__) || defined(__osf__) || \
5552 defined(__sgi) || defined(linux) || defined(_AIX51) || \
5553 (defined(sun) && !defined(__SVR4) && !defined(__svr4__))
5554 int unit, space;
5555 char temp[20];
5556 char *s;
5557 # endif
5558
5559 ASSERT(buffer != NULL);
5560 #ifdef notdef
5561 if (buffer == NULL)
5562 buffer = namebuf;
5563 #endif
5564 (void) strncpy(buffer, ifp->if_name, LIFNAMSIZ);
5565 buffer[LIFNAMSIZ - 1] = '\0';
5566 # if defined(MENTAT) || defined(__FreeBSD__) || defined(__osf__) || \
5567 defined(__sgi) || defined(_AIX51) || \
5568 (defined(sun) && !defined(__SVR4) && !defined(__svr4__))
5569 for (s = buffer; *s; s++)
5570 ;
5571 unit = ifp->if_unit;
5572 space = LIFNAMSIZ - (s - buffer);
5573 if (space > 0) {
5574 # if defined(SNPRINTF) && defined(_KERNEL)
5575 (void) SNPRINTF(temp, sizeof(temp), "%d", unit);
5576 # else
5577 (void) sprintf(temp, "%d", unit);
5578 # endif
5579 (void) strncpy(s, temp, space);
5580 }
5581 # endif
5582 return buffer;
5583 }
5584 #endif
5585
5586
5587 /* ------------------------------------------------------------------------ */
5588 /* Function: fr_ioctlswitch */
5589 /* Returns: int - -1 continue processing, else ioctl return value */
5590 /* Parameters: unit(I) - device unit opened */
5591 /* data(I) - pointer to ioctl data */
5592 /* cmd(I) - ioctl command */
5593 /* mode(I) - mode value */
5594 /* */
5595 /* Based on the value of unit, call the appropriate ioctl handler or return */
5596 /* EIO if ipfilter is not running. Also checks if write perms are req'd */
5597 /* for the device in order to execute the ioctl. */
5598 /* ------------------------------------------------------------------------ */
fr_ioctlswitch(unit,data,cmd,mode,uid,ctx,ifs)5599 INLINE int fr_ioctlswitch(unit, data, cmd, mode, uid, ctx, ifs)
5600 int unit, mode, uid;
5601 ioctlcmd_t cmd;
5602 void *data, *ctx;
5603 ipf_stack_t *ifs;
5604 {
5605 int error = 0;
5606
5607 switch (unit)
5608 {
5609 case IPL_LOGIPF :
5610 error = -1;
5611 break;
5612 case IPL_LOGNAT :
5613 if (ifs->ifs_fr_running > 0)
5614 error = fr_nat_ioctl(data, cmd, mode, uid, ctx, ifs);
5615 else
5616 error = EIO;
5617 break;
5618 case IPL_LOGSTATE :
5619 if (ifs->ifs_fr_running > 0)
5620 error = fr_state_ioctl(data, cmd, mode, uid, ctx, ifs);
5621 else
5622 error = EIO;
5623 break;
5624 case IPL_LOGAUTH :
5625 if (ifs->ifs_fr_running > 0) {
5626 if ((cmd == (ioctlcmd_t)SIOCADAFR) ||
5627 (cmd == (ioctlcmd_t)SIOCRMAFR)) {
5628 if (!(mode & FWRITE)) {
5629 error = EPERM;
5630 } else {
5631 error = frrequest(unit, cmd, data,
5632 ifs->ifs_fr_active, 1, ifs);
5633 }
5634 } else {
5635 error = fr_auth_ioctl(data, cmd, mode, uid, ctx, ifs);
5636 }
5637 } else
5638 error = EIO;
5639 break;
5640 case IPL_LOGSYNC :
5641 #ifdef IPFILTER_SYNC
5642 if (ifs->ifs_fr_running > 0)
5643 error = fr_sync_ioctl(data, cmd, mode, ifs);
5644 else
5645 #endif
5646 error = EIO;
5647 break;
5648 case IPL_LOGSCAN :
5649 #ifdef IPFILTER_SCAN
5650 if (ifs->ifs_fr_running > 0)
5651 error = fr_scan_ioctl(data, cmd, mode, ifs);
5652 else
5653 #endif
5654 error = EIO;
5655 break;
5656 case IPL_LOGLOOKUP :
5657 #ifdef IPFILTER_LOOKUP
5658 if (ifs->ifs_fr_running > 0)
5659 error = ip_lookup_ioctl(data, cmd, mode, uid, ctx, ifs);
5660 else
5661 #endif
5662 error = EIO;
5663 break;
5664 default :
5665 error = EIO;
5666 break;
5667 }
5668
5669 return error;
5670 }
5671
5672
5673 /*
5674 * This array defines the expected size of objects coming into the kernel
5675 * for the various recognised object types.
5676 */
5677 #define NUM_OBJ_TYPES 19
5678
5679 static int fr_objbytes[NUM_OBJ_TYPES][2] = {
5680 { 1, sizeof(struct frentry) }, /* frentry */
5681 { 0, sizeof(struct friostat) },
5682 { 0, sizeof(struct fr_info) },
5683 { 0, sizeof(struct fr_authstat) },
5684 { 0, sizeof(struct ipfrstat) },
5685 { 0, sizeof(struct ipnat) },
5686 { 0, sizeof(struct natstat) },
5687 { 0, sizeof(struct ipstate_save) },
5688 { 1, sizeof(struct nat_save) }, /* nat_save */
5689 { 0, sizeof(struct natlookup) },
5690 { 1, sizeof(struct ipstate) }, /* ipstate */
5691 { 0, sizeof(struct ips_stat) },
5692 { 0, sizeof(struct frauth) },
5693 { 0, sizeof(struct ipftune) },
5694 { 0, sizeof(struct nat) }, /* nat_t */
5695 { 0, sizeof(struct ipfruleiter) },
5696 { 0, sizeof(struct ipfgeniter) },
5697 { 0, sizeof(struct ipftable) },
5698 { 0, sizeof(struct ipflookupiter) }
5699 };
5700
5701
5702 /* ------------------------------------------------------------------------ */
5703 /* Function: fr_getzoneid */
5704 /* Returns: int - 0 = success, else failure */
5705 /* Parameters: idsp(I) - pointer to ipf_devstate_t */
5706 /* data(I) - pointer to ioctl data */
5707 /* */
5708 /* Set the zone ID in idsp based on the zone name in ipfzoneobj. Further */
5709 /* ioctls will act on the IPF stack for that zone ID. */
5710 /* ------------------------------------------------------------------------ */
5711 #if defined(_KERNEL)
fr_setzoneid(idsp,data)5712 int fr_setzoneid(idsp, data)
5713 ipf_devstate_t *idsp;
5714 void *data;
5715 {
5716 int error = 0;
5717 ipfzoneobj_t ipfzo;
5718 zone_t *zone;
5719
5720 error = BCOPYIN(data, &ipfzo, sizeof(ipfzo));
5721 if (error != 0)
5722 return EFAULT;
5723
5724 if (memchr(ipfzo.ipfz_zonename, '\0', ZONENAME_MAX) == NULL)
5725 return EFAULT;
5726
5727 /*
5728 * The global zone doesn't have a GZ-controlled stack, so no
5729 * sense in going any further
5730 */
5731 if (strcmp(ipfzo.ipfz_zonename, "global") == 0)
5732 return ENODEV;
5733
5734 if ((zone = zone_find_by_name(ipfzo.ipfz_zonename)) == NULL)
5735 return ENODEV;
5736
5737 /*
5738 * Store the zone ID that to control, and whether it's the
5739 * GZ-controlled stack that's wanted
5740 */
5741 idsp->ipfs_zoneid = zone->zone_id;
5742 idsp->ipfs_gz = (ipfzo.ipfz_gz == 1) ? B_TRUE : B_FALSE;
5743 zone_rele(zone);
5744
5745 return error;
5746 }
5747 #endif
5748
5749
5750 /* ------------------------------------------------------------------------ */
5751 /* Function: fr_inobj */
5752 /* Returns: int - 0 = success, else failure */
5753 /* Parameters: data(I) - pointer to ioctl data */
5754 /* ptr(I) - pointer to store real data in */
5755 /* type(I) - type of structure being moved */
5756 /* */
5757 /* Copy in the contents of what the ipfobj_t points to. In future, we */
5758 /* add things to check for version numbers, sizes, etc, to make it backward */
5759 /* compatible at the ABI for user land. */
5760 /* ------------------------------------------------------------------------ */
fr_inobj(data,ptr,type)5761 int fr_inobj(data, ptr, type)
5762 void *data;
5763 void *ptr;
5764 int type;
5765 {
5766 ipfobj_t obj;
5767 int error = 0;
5768
5769 if ((type < 0) || (type > NUM_OBJ_TYPES-1))
5770 return EINVAL;
5771
5772 error = BCOPYIN((caddr_t)data, (caddr_t)&obj, sizeof(obj));
5773 if (error != 0)
5774 return EFAULT;
5775
5776 if (obj.ipfo_type != type)
5777 return EINVAL;
5778
5779 #ifndef IPFILTER_COMPAT
5780 if ((fr_objbytes[type][0] & 1) != 0) {
5781 if (obj.ipfo_size < fr_objbytes[type][1])
5782 return EINVAL;
5783 } else if (obj.ipfo_size != fr_objbytes[type][1])
5784 return EINVAL;
5785 #else
5786 if (obj.ipfo_rev != IPFILTER_VERSION) {
5787 error = fr_incomptrans(&obj, ptr);
5788 return error;
5789 }
5790
5791 if ((fr_objbytes[type][0] & 1) != 0 &&
5792 obj.ipfo_size < fr_objbytes[type][1] ||
5793 obj.ipfo_size != fr_objbytes[type][1])
5794 return EINVAL;
5795 #endif
5796
5797 if ((fr_objbytes[type][0] & 1) != 0) {
5798 error = COPYIN((caddr_t)obj.ipfo_ptr, (caddr_t)ptr,
5799 fr_objbytes[type][1]);
5800 } else {
5801 error = COPYIN((caddr_t)obj.ipfo_ptr, (caddr_t)ptr,
5802 obj.ipfo_size);
5803 }
5804 return error;
5805 }
5806
5807
5808 /* ------------------------------------------------------------------------ */
5809 /* Function: fr_inobjsz */
5810 /* Returns: int - 0 = success, else failure */
5811 /* Parameters: data(I) - pointer to ioctl data */
5812 /* ptr(I) - pointer to store real data in */
5813 /* type(I) - type of structure being moved */
5814 /* sz(I) - size of data to copy */
5815 /* */
5816 /* As per fr_inobj, except the size of the object to copy in is passed in */
5817 /* but it must not be smaller than the size defined for the type and the */
5818 /* type must allow for varied sized objects. The extra requirement here is */
5819 /* that sz must match the size of the object being passed in - this is not */
5820 /* not possible nor required in fr_inobj(). */
5821 /* ------------------------------------------------------------------------ */
fr_inobjsz(data,ptr,type,sz)5822 int fr_inobjsz(data, ptr, type, sz)
5823 void *data;
5824 void *ptr;
5825 int type, sz;
5826 {
5827 ipfobj_t obj;
5828 int error;
5829
5830 if ((type < 0) || (type > NUM_OBJ_TYPES-1))
5831 return EINVAL;
5832 if (((fr_objbytes[type][0] & 1) == 0) || (sz < fr_objbytes[type][1]))
5833 return EINVAL;
5834
5835 error = BCOPYIN((caddr_t)data, (caddr_t)&obj, sizeof(obj));
5836 if (error != 0)
5837 return EFAULT;
5838
5839 if (obj.ipfo_type != type)
5840 return EINVAL;
5841
5842 #ifndef IPFILTER_COMPAT
5843 if (obj.ipfo_size != sz)
5844 return EINVAL;
5845 #else
5846 if (obj.ipfo_rev != IPFILTER_VERSION)
5847 /*XXX compatibility hook here */
5848 /*EMPTY*/;
5849 if (obj.ipfo_size != sz)
5850 /* XXX compatibility hook here */
5851 return EINVAL;
5852 #endif
5853
5854 error = COPYIN((caddr_t)obj.ipfo_ptr, (caddr_t)ptr, sz);
5855 return error;
5856 }
5857
5858
5859 /* ------------------------------------------------------------------------ */
5860 /* Function: fr_outobjsz */
5861 /* Returns: int - 0 = success, else failure */
5862 /* Parameters: data(I) - pointer to ioctl data */
5863 /* ptr(I) - pointer to store real data in */
5864 /* type(I) - type of structure being moved */
5865 /* sz(I) - size of data to copy */
5866 /* */
5867 /* As per fr_outobj, except the size of the object to copy out is passed in */
5868 /* but it must not be smaller than the size defined for the type and the */
5869 /* type must allow for varied sized objects. The extra requirement here is */
5870 /* that sz must match the size of the object being passed in - this is not */
5871 /* not possible nor required in fr_outobj(). */
5872 /* ------------------------------------------------------------------------ */
fr_outobjsz(data,ptr,type,sz)5873 int fr_outobjsz(data, ptr, type, sz)
5874 void *data;
5875 void *ptr;
5876 int type, sz;
5877 {
5878 ipfobj_t obj;
5879 int error;
5880
5881 if ((type < 0) || (type > NUM_OBJ_TYPES-1) ||
5882 ((fr_objbytes[type][0] & 1) == 0) ||
5883 (sz < fr_objbytes[type][1]))
5884 return EINVAL;
5885
5886 error = BCOPYIN((caddr_t)data, (caddr_t)&obj, sizeof(obj));
5887 if (error != 0)
5888 return EFAULT;
5889
5890 if (obj.ipfo_type != type)
5891 return EINVAL;
5892
5893 #ifndef IPFILTER_COMPAT
5894 if (obj.ipfo_size != sz)
5895 return EINVAL;
5896 #else
5897 if (obj.ipfo_rev != IPFILTER_VERSION)
5898 /* XXX compatibility hook here */
5899 /*EMPTY*/;
5900 if (obj.ipfo_size != sz)
5901 /* XXX compatibility hook here */
5902 return EINVAL;
5903 #endif
5904
5905 error = COPYOUT((caddr_t)ptr, (caddr_t)obj.ipfo_ptr, sz);
5906 return error;
5907 }
5908
5909
5910 /* ------------------------------------------------------------------------ */
5911 /* Function: fr_outobj */
5912 /* Returns: int - 0 = success, else failure */
5913 /* Parameters: data(I) - pointer to ioctl data */
5914 /* ptr(I) - pointer to store real data in */
5915 /* type(I) - type of structure being moved */
5916 /* */
5917 /* Copy out the contents of what ptr is to where ipfobj points to. In */
5918 /* future, we add things to check for version numbers, sizes, etc, to make */
5919 /* it backward compatible at the ABI for user land. */
5920 /* ------------------------------------------------------------------------ */
fr_outobj(data,ptr,type)5921 int fr_outobj(data, ptr, type)
5922 void *data;
5923 void *ptr;
5924 int type;
5925 {
5926 ipfobj_t obj;
5927 int error;
5928
5929 if ((type < 0) || (type > NUM_OBJ_TYPES-1))
5930 return EINVAL;
5931
5932 error = BCOPYIN((caddr_t)data, (caddr_t)&obj, sizeof(obj));
5933 if (error != 0)
5934 return EFAULT;
5935
5936 if (obj.ipfo_type != type)
5937 return EINVAL;
5938
5939 #ifndef IPFILTER_COMPAT
5940 if ((fr_objbytes[type][0] & 1) != 0) {
5941 if (obj.ipfo_size < fr_objbytes[type][1])
5942 return EINVAL;
5943 } else if (obj.ipfo_size != fr_objbytes[type][1])
5944 return EINVAL;
5945 #else
5946 if (obj.ipfo_rev != IPFILTER_VERSION) {
5947 error = fr_outcomptrans(&obj, ptr);
5948 return error;
5949 }
5950
5951 if ((fr_objbytes[type][0] & 1) != 0 &&
5952 obj.ipfo_size < fr_objbytes[type][1] ||
5953 obj.ipfo_size != fr_objbytes[type][1])
5954 return EINVAL;
5955 #endif
5956
5957 error = COPYOUT((caddr_t)ptr, (caddr_t)obj.ipfo_ptr, obj.ipfo_size);
5958 return error;
5959 }
5960
5961
5962 /* ------------------------------------------------------------------------ */
5963 /* Function: fr_checkl4sum */
5964 /* Returns: int - 0 = good, -1 = bad, 1 = cannot check */
5965 /* Parameters: fin(I) - pointer to packet information */
5966 /* */
5967 /* If possible, calculate the layer 4 checksum for the packet. If this is */
5968 /* not possible, return without indicating a failure or success but in a */
5969 /* way that is ditinguishable. */
5970 /* ------------------------------------------------------------------------ */
fr_checkl4sum(fin)5971 int fr_checkl4sum(fin)
5972 fr_info_t *fin;
5973 {
5974 u_short sum, hdrsum, *csump;
5975 udphdr_t *udp;
5976 int dosum;
5977 ipf_stack_t *ifs = fin->fin_ifs;
5978
5979 #if defined(SOLARIS) && defined(_KERNEL) && (SOLARIS2 >= 6)
5980 net_handle_t net_data_p;
5981 if (fin->fin_v == 4)
5982 net_data_p = ifs->ifs_ipf_ipv4;
5983 else
5984 net_data_p = ifs->ifs_ipf_ipv6;
5985 #endif
5986
5987 if ((fin->fin_flx & FI_NOCKSUM) != 0)
5988 return 0;
5989
5990 /*
5991 * If the TCP packet isn't a fragment, isn't too short and otherwise
5992 * isn't already considered "bad", then validate the checksum. If
5993 * this check fails then considered the packet to be "bad".
5994 */
5995 if ((fin->fin_flx & (FI_FRAG|FI_SHORT|FI_BAD)) != 0)
5996 return 1;
5997
5998 csump = NULL;
5999 hdrsum = 0;
6000 dosum = 0;
6001 sum = 0;
6002
6003 #if defined(SOLARIS) && defined(_KERNEL) && (SOLARIS2 >= 6)
6004 ASSERT(fin->fin_m != NULL);
6005 if (NET_IS_HCK_L4_FULL(net_data_p, fin->fin_m) ||
6006 NET_IS_HCK_L4_PART(net_data_p, fin->fin_m)) {
6007 hdrsum = 0;
6008 sum = 0;
6009 } else {
6010 #endif
6011 switch (fin->fin_p)
6012 {
6013 case IPPROTO_TCP :
6014 csump = &((tcphdr_t *)fin->fin_dp)->th_sum;
6015 dosum = 1;
6016 break;
6017
6018 case IPPROTO_UDP :
6019 udp = fin->fin_dp;
6020 if (udp->uh_sum != 0) {
6021 csump = &udp->uh_sum;
6022 dosum = 1;
6023 }
6024 break;
6025
6026 case IPPROTO_ICMP :
6027 csump = &((struct icmp *)fin->fin_dp)->icmp_cksum;
6028 dosum = 1;
6029 break;
6030
6031 default :
6032 return 1;
6033 /*NOTREACHED*/
6034 }
6035
6036 if (csump != NULL)
6037 hdrsum = *csump;
6038
6039 if (dosum)
6040 sum = fr_cksum(fin->fin_m, fin->fin_ip,
6041 fin->fin_p, fin->fin_dp);
6042 #if defined(SOLARIS) && defined(_KERNEL) && (SOLARIS2 >= 6)
6043 }
6044 #endif
6045 #if !defined(_KERNEL)
6046 if (sum == hdrsum) {
6047 FR_DEBUG(("checkl4sum: %hx == %hx\n", sum, hdrsum));
6048 } else {
6049 FR_DEBUG(("checkl4sum: %hx != %hx\n", sum, hdrsum));
6050 }
6051 #endif
6052 if (hdrsum == sum)
6053 return 0;
6054 return -1;
6055 }
6056
6057
6058 /* ------------------------------------------------------------------------ */
6059 /* Function: fr_ifpfillv4addr */
6060 /* Returns: int - 0 = address update, -1 = address not updated */
6061 /* Parameters: atype(I) - type of network address update to perform */
6062 /* sin(I) - pointer to source of address information */
6063 /* mask(I) - pointer to source of netmask information */
6064 /* inp(I) - pointer to destination address store */
6065 /* inpmask(I) - pointer to destination netmask store */
6066 /* */
6067 /* Given a type of network address update (atype) to perform, copy */
6068 /* information from sin/mask into inp/inpmask. If ipnmask is NULL then no */
6069 /* netmask update is performed unless FRI_NETMASKED is passed as atype, in */
6070 /* which case the operation fails. For all values of atype other than */
6071 /* FRI_NETMASKED, if inpmask is non-NULL then the mask is set to an all 1s */
6072 /* value. */
6073 /* ------------------------------------------------------------------------ */
fr_ifpfillv4addr(atype,sin,mask,inp,inpmask)6074 int fr_ifpfillv4addr(atype, sin, mask, inp, inpmask)
6075 int atype;
6076 struct sockaddr_in *sin, *mask;
6077 struct in_addr *inp, *inpmask;
6078 {
6079 if (inpmask != NULL && atype != FRI_NETMASKED)
6080 inpmask->s_addr = 0xffffffff;
6081
6082 if (atype == FRI_NETWORK || atype == FRI_NETMASKED) {
6083 if (atype == FRI_NETMASKED) {
6084 if (inpmask == NULL)
6085 return -1;
6086 inpmask->s_addr = mask->sin_addr.s_addr;
6087 }
6088 inp->s_addr = sin->sin_addr.s_addr & mask->sin_addr.s_addr;
6089 } else {
6090 inp->s_addr = sin->sin_addr.s_addr;
6091 }
6092 return 0;
6093 }
6094
6095
6096 #ifdef USE_INET6
6097 /* ------------------------------------------------------------------------ */
6098 /* Function: fr_ifpfillv6addr */
6099 /* Returns: int - 0 = address update, -1 = address not updated */
6100 /* Parameters: atype(I) - type of network address update to perform */
6101 /* sin(I) - pointer to source of address information */
6102 /* mask(I) - pointer to source of netmask information */
6103 /* inp(I) - pointer to destination address store */
6104 /* inpmask(I) - pointer to destination netmask store */
6105 /* */
6106 /* Given a type of network address update (atype) to perform, copy */
6107 /* information from sin/mask into inp/inpmask. If ipnmask is NULL then no */
6108 /* netmask update is performed unless FRI_NETMASKED is passed as atype, in */
6109 /* which case the operation fails. For all values of atype other than */
6110 /* FRI_NETMASKED, if inpmask is non-NULL then the mask is set to an all 1s */
6111 /* value. */
6112 /* ------------------------------------------------------------------------ */
fr_ifpfillv6addr(atype,sin,mask,inp,inpmask)6113 int fr_ifpfillv6addr(atype, sin, mask, inp, inpmask)
6114 int atype;
6115 struct sockaddr_in6 *sin, *mask;
6116 struct in_addr *inp, *inpmask;
6117 {
6118 i6addr_t *src, *dst, *and, *dmask;
6119
6120 src = (i6addr_t *)&sin->sin6_addr;
6121 and = (i6addr_t *)&mask->sin6_addr;
6122 dst = (i6addr_t *)inp;
6123 dmask = (i6addr_t *)inpmask;
6124
6125 if (inpmask != NULL && atype != FRI_NETMASKED) {
6126 dmask->i6[0] = 0xffffffff;
6127 dmask->i6[1] = 0xffffffff;
6128 dmask->i6[2] = 0xffffffff;
6129 dmask->i6[3] = 0xffffffff;
6130 }
6131
6132 if (atype == FRI_NETWORK || atype == FRI_NETMASKED) {
6133 if (atype == FRI_NETMASKED) {
6134 if (inpmask == NULL)
6135 return -1;
6136 dmask->i6[0] = and->i6[0];
6137 dmask->i6[1] = and->i6[1];
6138 dmask->i6[2] = and->i6[2];
6139 dmask->i6[3] = and->i6[3];
6140 }
6141
6142 dst->i6[0] = src->i6[0] & and->i6[0];
6143 dst->i6[1] = src->i6[1] & and->i6[1];
6144 dst->i6[2] = src->i6[2] & and->i6[2];
6145 dst->i6[3] = src->i6[3] & and->i6[3];
6146 } else {
6147 dst->i6[0] = src->i6[0];
6148 dst->i6[1] = src->i6[1];
6149 dst->i6[2] = src->i6[2];
6150 dst->i6[3] = src->i6[3];
6151 }
6152 return 0;
6153 }
6154 #endif
6155
6156
6157 /* ------------------------------------------------------------------------ */
6158 /* Function: fr_matchtag */
6159 /* Returns: 0 == mismatch, 1 == match. */
6160 /* Parameters: tag1(I) - pointer to first tag to compare */
6161 /* tag2(I) - pointer to second tag to compare */
6162 /* */
6163 /* Returns true (non-zero) or false(0) if the two tag structures can be */
6164 /* considered to be a match or not match, respectively. The tag is 16 */
6165 /* bytes long (16 characters) but that is overlayed with 4 32bit ints so */
6166 /* compare the ints instead, for speed. tag1 is the master of the */
6167 /* comparison. This function should only be called with both tag1 and tag2 */
6168 /* as non-NULL pointers. */
6169 /* ------------------------------------------------------------------------ */
fr_matchtag(tag1,tag2)6170 int fr_matchtag(tag1, tag2)
6171 ipftag_t *tag1, *tag2;
6172 {
6173 if (tag1 == tag2)
6174 return 1;
6175
6176 if ((tag1->ipt_num[0] == 0) && (tag2->ipt_num[0] == 0))
6177 return 1;
6178
6179 if ((tag1->ipt_num[0] == tag2->ipt_num[0]) &&
6180 (tag1->ipt_num[1] == tag2->ipt_num[1]) &&
6181 (tag1->ipt_num[2] == tag2->ipt_num[2]) &&
6182 (tag1->ipt_num[3] == tag2->ipt_num[3]))
6183 return 1;
6184 return 0;
6185 }
6186
6187
6188 /* ------------------------------------------------------------------------ */
6189 /* Function: fr_coalesce */
6190 /* Returns: 1 == success, -1 == failure, 0 == no change */
6191 /* Parameters: fin(I) - pointer to packet information */
6192 /* */
6193 /* Attempt to get all of the packet data into a single, contiguous buffer. */
6194 /* If this call returns a failure then the buffers have also been freed. */
6195 /* ------------------------------------------------------------------------ */
fr_coalesce(fin)6196 int fr_coalesce(fin)
6197 fr_info_t *fin;
6198 {
6199 ipf_stack_t *ifs = fin->fin_ifs;
6200 if ((fin->fin_flx & FI_COALESCE) != 0)
6201 return 1;
6202
6203 /*
6204 * If the mbuf pointers indicate that there is no mbuf to work with,
6205 * return but do not indicate success or failure.
6206 */
6207 if (fin->fin_m == NULL || fin->fin_mp == NULL)
6208 return 0;
6209
6210 #if defined(_KERNEL)
6211 if (fr_pullup(fin->fin_m, fin, fin->fin_plen) == NULL) {
6212 IPF_BUMP(ifs->ifs_fr_badcoalesces[fin->fin_out]);
6213 # ifdef MENTAT
6214 FREE_MB_T(*fin->fin_mp);
6215 # endif
6216 *fin->fin_mp = NULL;
6217 fin->fin_m = NULL;
6218 return -1;
6219 }
6220 #else
6221 fin = fin; /* LINT */
6222 #endif
6223 return 1;
6224 }
6225
6226
6227 /*
6228 * The following table lists all of the tunable variables that can be
6229 * accessed via SIOCIPFGET/SIOCIPFSET/SIOCIPFGETNEXT. The format of each row
6230 * in the table below is as follows:
6231 *
6232 * pointer to value, name of value, minimum, maximum, size of the value's
6233 * container, value attribute flags
6234 *
6235 * For convienience, IPFT_RDONLY means the value is read-only, IPFT_WRDISABLED
6236 * means the value can only be written to when IPFilter is loaded but disabled.
6237 * The obvious implication is if neither of these are set then the value can be
6238 * changed at any time without harm.
6239 */
6240 ipftuneable_t lcl_ipf_tuneables[] = {
6241 /* filtering */
6242 { { NULL }, "fr_flags", 0, 0xffffffff,
6243 0, 0 },
6244 { { NULL }, "fr_active", 0, 0,
6245 0, IPFT_RDONLY },
6246 { { NULL }, "fr_control_forwarding", 0, 1,
6247 0, 0 },
6248 { { NULL }, "fr_update_ipid", 0, 1,
6249 0, 0 },
6250 { { NULL }, "fr_chksrc", 0, 1,
6251 0, 0 },
6252 { { NULL }, "fr_minttl", 0, 1,
6253 0, 0 },
6254 { { NULL }, "fr_icmpminfragmtu", 0, 1,
6255 0, 0 },
6256 { { NULL }, "fr_pass", 0, 0xffffffff,
6257 0, 0 },
6258 #if SOLARIS2 >= 10
6259 { { NULL }, "ipf_loopback", 0, 1,
6260 0, IPFT_WRDISABLED },
6261 #endif
6262 /* state */
6263 { { NULL }, "fr_tcpidletimeout", 1, 0x7fffffff,
6264 0, IPFT_WRDISABLED },
6265 { { NULL }, "fr_tcpclosewait", 1, 0x7fffffff,
6266 0, IPFT_WRDISABLED },
6267 { { NULL }, "fr_tcplastack", 1, 0x7fffffff,
6268 0, IPFT_WRDISABLED },
6269 { { NULL }, "fr_tcptimeout", 1, 0x7fffffff,
6270 0, IPFT_WRDISABLED },
6271 { { NULL }, "fr_tcpclosed", 1, 0x7fffffff,
6272 0, IPFT_WRDISABLED },
6273 { { NULL }, "fr_tcphalfclosed", 1, 0x7fffffff,
6274 0, IPFT_WRDISABLED },
6275 { { NULL }, "fr_udptimeout", 1, 0x7fffffff,
6276 0, IPFT_WRDISABLED },
6277 { { NULL }, "fr_udpacktimeout", 1, 0x7fffffff,
6278 0, IPFT_WRDISABLED },
6279 { { NULL }, "fr_icmptimeout", 1, 0x7fffffff,
6280 0, IPFT_WRDISABLED },
6281 { { NULL }, "fr_icmpacktimeout", 1, 0x7fffffff,
6282 0, IPFT_WRDISABLED },
6283 { { NULL }, "fr_iptimeout", 1, 0x7fffffff,
6284 0, IPFT_WRDISABLED },
6285 { { NULL }, "fr_statemax", 1, 0x7fffffff,
6286 0, 0 },
6287 { { NULL }, "fr_statesize", 1, 0x7fffffff,
6288 0, IPFT_WRDISABLED },
6289 { { NULL }, "fr_state_lock", 0, 1,
6290 0, IPFT_RDONLY },
6291 { { NULL }, "fr_state_maxbucket", 1, 0x7fffffff,
6292 0, IPFT_WRDISABLED },
6293 { { NULL }, "fr_state_maxbucket_reset", 0, 1,
6294 0, IPFT_WRDISABLED },
6295 { { NULL }, "ipstate_logging", 0, 1,
6296 0, 0 },
6297 { { NULL }, "state_flush_level_hi", 1, 100,
6298 0, 0 },
6299 { { NULL }, "state_flush_level_lo", 1, 100,
6300 0, 0 },
6301 /* nat */
6302 { { NULL }, "fr_nat_lock", 0, 1,
6303 0, IPFT_RDONLY },
6304 { { NULL }, "ipf_nattable_sz", 1, 0x7fffffff,
6305 0, IPFT_WRDISABLED },
6306 { { NULL }, "ipf_nattable_max", 1, 0x7fffffff,
6307 0, 0 },
6308 { { NULL }, "ipf_natrules_sz", 1, 0x7fffffff,
6309 0, IPFT_WRDISABLED },
6310 { { NULL }, "ipf_rdrrules_sz", 1, 0x7fffffff,
6311 0, IPFT_WRDISABLED },
6312 { { NULL }, "ipf_hostmap_sz", 1, 0x7fffffff,
6313 0, IPFT_WRDISABLED },
6314 { { NULL }, "fr_nat_maxbucket", 1, 0x7fffffff,
6315 0, IPFT_WRDISABLED },
6316 { { NULL }, "fr_nat_maxbucket_reset", 0, 1,
6317 0, IPFT_WRDISABLED },
6318 { { NULL }, "nat_logging", 0, 1,
6319 0, 0 },
6320 { { NULL }, "fr_defnatage", 1, 0x7fffffff,
6321 0, IPFT_WRDISABLED },
6322 { { NULL }, "fr_defnatipage", 1, 0x7fffffff,
6323 0, IPFT_WRDISABLED },
6324 { { NULL }, "fr_defnaticmpage", 1, 0x7fffffff,
6325 0, IPFT_WRDISABLED },
6326 { { NULL }, "nat_flush_level_hi", 1, 100,
6327 0, 0 },
6328 { { NULL }, "nat_flush_level_lo", 1, 100,
6329 0, 0 },
6330 /* frag */
6331 { { NULL }, "ipfr_size", 1, 0x7fffffff,
6332 0, IPFT_WRDISABLED },
6333 { { NULL }, "fr_ipfrttl", 1, 0x7fffffff,
6334 0, IPFT_WRDISABLED },
6335 #ifdef IPFILTER_LOG
6336 /* log */
6337 { { NULL }, "ipl_suppress", 0, 1,
6338 0, 0 },
6339 { { NULL }, "ipl_buffer_sz", 0, 0,
6340 0, IPFT_RDONLY },
6341 { { NULL }, "ipl_logmax", 0, 0x7fffffff,
6342 0, IPFT_WRDISABLED },
6343 { { NULL }, "ipl_logall", 0, 1,
6344 0, 0 },
6345 { { NULL }, "ipl_logsize", 0, 0x80000,
6346 0, 0 },
6347 #endif
6348 { { NULL }, NULL, 0, 0 }
6349 };
6350
6351 static ipftuneable_t *
tune_lookup(ipf_stack_t * ifs,char * name)6352 tune_lookup(ipf_stack_t *ifs, char *name)
6353 {
6354 int i;
6355
6356 for (i = 0; ifs->ifs_ipf_tuneables[i].ipft_name != NULL; i++) {
6357 if (strcmp(ifs->ifs_ipf_tuneables[i].ipft_name, name) == 0)
6358 return (&ifs->ifs_ipf_tuneables[i]);
6359 }
6360 return (NULL);
6361 }
6362
6363 #ifdef _KERNEL
6364 extern dev_info_t *ipf_dev_info;
6365 extern int ipf_property_update __P((dev_info_t *, ipf_stack_t *));
6366 #endif
6367
6368 /* -------------------------------------------------------------------- */
6369 /* Function: ipftuneable_setdefs() */
6370 /* Returns: void */
6371 /* Parameters: ifs - pointer to newly allocated IPF instance */
6372 /* assigned to IP instance */
6373 /* */
6374 /* Function initializes IPF instance variables. Function is invoked */
6375 /* from ipftuneable_alloc(). ipftuneable_alloc() is called only one */
6376 /* time during IP instance lifetime - at the time of IP instance */
6377 /* creation. Anytime IP instance is being created new private IPF */
6378 /* instance is allocated and assigned to it. The moment of IP */
6379 /* instance creation is the right time to initialize those IPF */
6380 /* variables. */
6381 /* */
6382 /* -------------------------------------------------------------------- */
ipftuneable_setdefs(ipf_stack_t * ifs)6383 static void ipftuneable_setdefs(ipf_stack_t *ifs)
6384 {
6385 ifs->ifs_ipfr_size = IPFT_SIZE;
6386 ifs->ifs_fr_ipfrttl = 120; /* 60 seconds */
6387
6388 /* it comes from fr_authinit() in IPF auth */
6389 ifs->ifs_fr_authsize = FR_NUMAUTH;
6390 ifs->ifs_fr_defaultauthage = 600;
6391
6392 /* it comes from fr_stateinit() in IPF state */
6393 ifs->ifs_fr_tcpidletimeout = IPF_TTLVAL(3600 * 24 * 5); /* five days */
6394 ifs->ifs_fr_tcpclosewait = IPF_TTLVAL(TCP_MSL);
6395 ifs->ifs_fr_tcplastack = IPF_TTLVAL(TCP_MSL);
6396 ifs->ifs_fr_tcptimeout = IPF_TTLVAL(TCP_MSL);
6397 ifs->ifs_fr_tcpclosed = IPF_TTLVAL(60);
6398 ifs->ifs_fr_tcphalfclosed = IPF_TTLVAL(2 * 3600); /* 2 hours */
6399 ifs->ifs_fr_udptimeout = IPF_TTLVAL(120);
6400 ifs->ifs_fr_udpacktimeout = IPF_TTLVAL(12);
6401 ifs->ifs_fr_icmptimeout = IPF_TTLVAL(60);
6402 ifs->ifs_fr_icmpacktimeout = IPF_TTLVAL(6);
6403 ifs->ifs_fr_iptimeout = IPF_TTLVAL(60);
6404 ifs->ifs_fr_statemax = IPSTATE_MAX;
6405 ifs->ifs_fr_statesize = IPSTATE_SIZE;
6406 ifs->ifs_fr_state_maxbucket_reset = 1;
6407 ifs->ifs_state_flush_level_hi = ST_FLUSH_HI;
6408 ifs->ifs_state_flush_level_lo = ST_FLUSH_LO;
6409
6410 /* it comes from fr_natinit() in ipnat */
6411 ifs->ifs_ipf_nattable_sz = NAT_TABLE_SZ;
6412 ifs->ifs_ipf_nattable_max = NAT_TABLE_MAX;
6413 ifs->ifs_ipf_natrules_sz = NAT_SIZE;
6414 ifs->ifs_ipf_rdrrules_sz = RDR_SIZE;
6415 ifs->ifs_ipf_hostmap_sz = HOSTMAP_SIZE;
6416 ifs->ifs_fr_nat_maxbucket_reset = 1;
6417 ifs->ifs_fr_defnatage = DEF_NAT_AGE;
6418 ifs->ifs_fr_defnatipage = 120; /* 60 seconds */
6419 ifs->ifs_fr_defnaticmpage = 6; /* 3 seconds */
6420 ifs->ifs_nat_flush_level_hi = NAT_FLUSH_HI;
6421 ifs->ifs_nat_flush_level_lo = NAT_FLUSH_LO;
6422
6423 #ifdef IPFILTER_LOG
6424 /* it comes from fr_loginit() in IPF log */
6425 ifs->ifs_ipl_suppress = 1;
6426 ifs->ifs_ipl_logmax = IPL_LOGMAX;
6427 ifs->ifs_ipl_logsize = IPFILTER_LOGSIZE;
6428
6429 /* from fr_natinit() */
6430 ifs->ifs_nat_logging = 1;
6431
6432 /* from fr_stateinit() */
6433 ifs->ifs_ipstate_logging = 1;
6434 #else
6435 /* from fr_natinit() */
6436 ifs->ifs_nat_logging = 0;
6437
6438 /* from fr_stateinit() */
6439 ifs->ifs_ipstate_logging = 0;
6440 #endif
6441 ifs->ifs_ipf_loopback = 0;
6442
6443 }
6444 /*
6445 * Allocate a per-stack tuneable and copy in the names. Then
6446 * set it to point to each of the per-stack tunables.
6447 */
6448 void
ipftuneable_alloc(ipf_stack_t * ifs)6449 ipftuneable_alloc(ipf_stack_t *ifs)
6450 {
6451 ipftuneable_t *item;
6452
6453 /*
6454 * We are being called as part of netstack creation and may not return
6455 * NULL; use a sleeping allocation.
6456 */
6457 SLEEPING_KMALLOCS(ifs->ifs_ipf_tuneables, ipftuneable_t *,
6458 sizeof (lcl_ipf_tuneables));
6459 bcopy(lcl_ipf_tuneables, ifs->ifs_ipf_tuneables,
6460 sizeof (lcl_ipf_tuneables));
6461
6462 #define TUNE_SET(_ifs, _name, _field) \
6463 item = tune_lookup((_ifs), (_name)); \
6464 if (item != NULL) { \
6465 item->ipft_una.ipftp_int = (unsigned int *)&((_ifs)->_field); \
6466 item->ipft_sz = sizeof ((_ifs)->_field); \
6467 }
6468
6469 TUNE_SET(ifs, "fr_flags", ifs_fr_flags);
6470 TUNE_SET(ifs, "fr_active", ifs_fr_active);
6471 TUNE_SET(ifs, "fr_control_forwarding", ifs_fr_control_forwarding);
6472 TUNE_SET(ifs, "fr_update_ipid", ifs_fr_update_ipid);
6473 TUNE_SET(ifs, "fr_chksrc", ifs_fr_chksrc);
6474 TUNE_SET(ifs, "fr_minttl", ifs_fr_minttl);
6475 TUNE_SET(ifs, "fr_icmpminfragmtu", ifs_fr_icmpminfragmtu);
6476 TUNE_SET(ifs, "fr_pass", ifs_fr_pass);
6477 TUNE_SET(ifs, "fr_tcpidletimeout", ifs_fr_tcpidletimeout);
6478 TUNE_SET(ifs, "fr_tcpclosewait", ifs_fr_tcpclosewait);
6479 TUNE_SET(ifs, "fr_tcplastack", ifs_fr_tcplastack);
6480 TUNE_SET(ifs, "fr_tcptimeout", ifs_fr_tcptimeout);
6481 TUNE_SET(ifs, "fr_tcpclosed", ifs_fr_tcpclosed);
6482 TUNE_SET(ifs, "fr_tcphalfclosed", ifs_fr_tcphalfclosed);
6483 TUNE_SET(ifs, "fr_udptimeout", ifs_fr_udptimeout);
6484 TUNE_SET(ifs, "fr_udpacktimeout", ifs_fr_udpacktimeout);
6485 TUNE_SET(ifs, "fr_icmptimeout", ifs_fr_icmptimeout);
6486 TUNE_SET(ifs, "fr_icmpacktimeout", ifs_fr_icmpacktimeout);
6487 TUNE_SET(ifs, "fr_iptimeout", ifs_fr_iptimeout);
6488 TUNE_SET(ifs, "fr_statemax", ifs_fr_statemax);
6489 TUNE_SET(ifs, "fr_statesize", ifs_fr_statesize);
6490 TUNE_SET(ifs, "fr_state_lock", ifs_fr_state_lock);
6491 TUNE_SET(ifs, "fr_state_maxbucket", ifs_fr_state_maxbucket);
6492 TUNE_SET(ifs, "fr_state_maxbucket_reset", ifs_fr_state_maxbucket_reset);
6493 TUNE_SET(ifs, "ipstate_logging", ifs_ipstate_logging);
6494 TUNE_SET(ifs, "fr_nat_lock", ifs_fr_nat_lock);
6495 TUNE_SET(ifs, "ipf_nattable_sz", ifs_ipf_nattable_sz);
6496 TUNE_SET(ifs, "ipf_nattable_max", ifs_ipf_nattable_max);
6497 TUNE_SET(ifs, "ipf_natrules_sz", ifs_ipf_natrules_sz);
6498 TUNE_SET(ifs, "ipf_rdrrules_sz", ifs_ipf_rdrrules_sz);
6499 TUNE_SET(ifs, "ipf_hostmap_sz", ifs_ipf_hostmap_sz);
6500 TUNE_SET(ifs, "fr_nat_maxbucket", ifs_fr_nat_maxbucket);
6501 TUNE_SET(ifs, "fr_nat_maxbucket_reset", ifs_fr_nat_maxbucket_reset);
6502 TUNE_SET(ifs, "nat_logging", ifs_nat_logging);
6503 TUNE_SET(ifs, "fr_defnatage", ifs_fr_defnatage);
6504 TUNE_SET(ifs, "fr_defnatipage", ifs_fr_defnatipage);
6505 TUNE_SET(ifs, "fr_defnaticmpage", ifs_fr_defnaticmpage);
6506 TUNE_SET(ifs, "nat_flush_level_hi", ifs_nat_flush_level_hi);
6507 TUNE_SET(ifs, "nat_flush_level_lo", ifs_nat_flush_level_lo);
6508 TUNE_SET(ifs, "state_flush_level_hi", ifs_state_flush_level_hi);
6509 TUNE_SET(ifs, "state_flush_level_lo", ifs_state_flush_level_lo);
6510 TUNE_SET(ifs, "ipfr_size", ifs_ipfr_size);
6511 TUNE_SET(ifs, "fr_ipfrttl", ifs_fr_ipfrttl);
6512 TUNE_SET(ifs, "ipf_loopback", ifs_ipf_loopback);
6513 #ifdef IPFILTER_LOG
6514 TUNE_SET(ifs, "ipl_suppress", ifs_ipl_suppress);
6515 TUNE_SET(ifs, "ipl_buffer_sz", ifs_ipl_buffer_sz);
6516 TUNE_SET(ifs, "ipl_logmax", ifs_ipl_logmax);
6517 TUNE_SET(ifs, "ipl_logall", ifs_ipl_logall);
6518 TUNE_SET(ifs, "ipl_logsize", ifs_ipl_logsize);
6519 #endif
6520 #undef TUNE_SET
6521
6522 ipftuneable_setdefs(ifs);
6523
6524 #ifdef _KERNEL
6525 (void) ipf_property_update(ipf_dev_info, ifs);
6526 #endif
6527 }
6528
6529 void
ipftuneable_free(ipf_stack_t * ifs)6530 ipftuneable_free(ipf_stack_t *ifs)
6531 {
6532 KFREES(ifs->ifs_ipf_tuneables, sizeof (lcl_ipf_tuneables));
6533 ifs->ifs_ipf_tuneables = NULL;
6534 }
6535
6536 /* ------------------------------------------------------------------------ */
6537 /* Function: fr_findtunebycookie */
6538 /* Returns: NULL = search failed, else pointer to tune struct */
6539 /* Parameters: cookie(I) - cookie value to search for amongst tuneables */
6540 /* next(O) - pointer to place to store the cookie for the */
6541 /* "next" tuneable, if it is desired. */
6542 /* */
6543 /* This function is used to walk through all of the existing tunables with */
6544 /* successive calls. It searches the known tunables for the one which has */
6545 /* a matching value for "cookie" - ie its address. When returning a match, */
6546 /* the next one to be found may be returned inside next. */
6547 /* ------------------------------------------------------------------------ */
fr_findtunebycookie(cookie,next,ifs)6548 static ipftuneable_t *fr_findtunebycookie(cookie, next, ifs)
6549 void *cookie, **next;
6550 ipf_stack_t * ifs;
6551 {
6552 ipftuneable_t *ta, **tap;
6553
6554 for (ta = ifs->ifs_ipf_tuneables; ta->ipft_name != NULL; ta++)
6555 if (ta == cookie) {
6556 if (next != NULL) {
6557 /*
6558 * If the next entry in the array has a name
6559 * present, then return a pointer to it for
6560 * where to go next, else return a pointer to
6561 * the dynaminc list as a key to search there
6562 * next. This facilitates a weak linking of
6563 * the two "lists" together.
6564 */
6565 if ((ta + 1)->ipft_name != NULL)
6566 *next = ta + 1;
6567 else
6568 *next = &ifs->ifs_ipf_tunelist;
6569 }
6570 return ta;
6571 }
6572
6573 for (tap = &ifs->ifs_ipf_tunelist; (ta = *tap) != NULL; tap = &ta->ipft_next)
6574 if (tap == cookie) {
6575 if (next != NULL)
6576 *next = &ta->ipft_next;
6577 return ta;
6578 }
6579
6580 if (next != NULL)
6581 *next = NULL;
6582 return NULL;
6583 }
6584
6585
6586 /* ------------------------------------------------------------------------ */
6587 /* Function: fr_findtunebyname */
6588 /* Returns: NULL = search failed, else pointer to tune struct */
6589 /* Parameters: name(I) - name of the tuneable entry to find. */
6590 /* */
6591 /* Search the static array of tuneables and the list of dynamic tuneables */
6592 /* for an entry with a matching name. If we can find one, return a pointer */
6593 /* to the matching structure. */
6594 /* ------------------------------------------------------------------------ */
fr_findtunebyname(name,ifs)6595 static ipftuneable_t *fr_findtunebyname(name, ifs)
6596 const char *name;
6597 ipf_stack_t *ifs;
6598 {
6599 ipftuneable_t *ta;
6600
6601 for (ta = ifs->ifs_ipf_tuneables; ta->ipft_name != NULL; ta++)
6602 if (!strcmp(ta->ipft_name, name)) {
6603 return ta;
6604 }
6605
6606 for (ta = ifs->ifs_ipf_tunelist; ta != NULL; ta = ta->ipft_next)
6607 if (!strcmp(ta->ipft_name, name)) {
6608 return ta;
6609 }
6610
6611 return NULL;
6612 }
6613
6614
6615 /* ------------------------------------------------------------------------ */
6616 /* Function: fr_addipftune */
6617 /* Returns: int - 0 == success, else failure */
6618 /* Parameters: newtune - pointer to new tune struct to add to tuneables */
6619 /* */
6620 /* Appends the tune structure pointer to by "newtune" to the end of the */
6621 /* current list of "dynamic" tuneable parameters. Once added, the owner */
6622 /* of the object is not expected to ever change "ipft_next". */
6623 /* ------------------------------------------------------------------------ */
fr_addipftune(newtune,ifs)6624 int fr_addipftune(newtune, ifs)
6625 ipftuneable_t *newtune;
6626 ipf_stack_t *ifs;
6627 {
6628 ipftuneable_t *ta, **tap;
6629
6630 ta = fr_findtunebyname(newtune->ipft_name, ifs);
6631 if (ta != NULL)
6632 return EEXIST;
6633
6634 for (tap = &ifs->ifs_ipf_tunelist; *tap != NULL; tap = &(*tap)->ipft_next)
6635 ;
6636
6637 newtune->ipft_next = NULL;
6638 *tap = newtune;
6639 return 0;
6640 }
6641
6642
6643 /* ------------------------------------------------------------------------ */
6644 /* Function: fr_delipftune */
6645 /* Returns: int - 0 == success, else failure */
6646 /* Parameters: oldtune - pointer to tune struct to remove from the list of */
6647 /* current dynamic tuneables */
6648 /* */
6649 /* Search for the tune structure, by pointer, in the list of those that are */
6650 /* dynamically added at run time. If found, adjust the list so that this */
6651 /* structure is no longer part of it. */
6652 /* ------------------------------------------------------------------------ */
fr_delipftune(oldtune,ifs)6653 int fr_delipftune(oldtune, ifs)
6654 ipftuneable_t *oldtune;
6655 ipf_stack_t *ifs;
6656 {
6657 ipftuneable_t *ta, **tap;
6658
6659 for (tap = &ifs->ifs_ipf_tunelist; (ta = *tap) != NULL; tap = &ta->ipft_next)
6660 if (ta == oldtune) {
6661 *tap = oldtune->ipft_next;
6662 oldtune->ipft_next = NULL;
6663 return 0;
6664 }
6665
6666 return ESRCH;
6667 }
6668
6669
6670 /* ------------------------------------------------------------------------ */
6671 /* Function: fr_ipftune */
6672 /* Returns: int - 0 == success, else failure */
6673 /* Parameters: cmd(I) - ioctl command number */
6674 /* data(I) - pointer to ioctl data structure */
6675 /* */
6676 /* Implement handling of SIOCIPFGETNEXT, SIOCIPFGET and SIOCIPFSET. These */
6677 /* three ioctls provide the means to access and control global variables */
6678 /* within IPFilter, allowing (for example) timeouts and table sizes to be */
6679 /* changed without rebooting, reloading or recompiling. The initialisation */
6680 /* and 'destruction' routines of the various components of ipfilter are all */
6681 /* each responsible for handling their own values being too big. */
6682 /* ------------------------------------------------------------------------ */
fr_ipftune(cmd,data,ifs)6683 int fr_ipftune(cmd, data, ifs)
6684 ioctlcmd_t cmd;
6685 void *data;
6686 ipf_stack_t *ifs;
6687 {
6688 ipftuneable_t *ta;
6689 ipftune_t tu;
6690 void *cookie;
6691 int error;
6692
6693 error = fr_inobj(data, &tu, IPFOBJ_TUNEABLE);
6694 if (error != 0)
6695 return error;
6696
6697 tu.ipft_name[sizeof(tu.ipft_name) - 1] = '\0';
6698 cookie = tu.ipft_cookie;
6699 ta = NULL;
6700
6701 switch (cmd)
6702 {
6703 case SIOCIPFGETNEXT :
6704 /*
6705 * If cookie is non-NULL, assume it to be a pointer to the last
6706 * entry we looked at, so find it (if possible) and return a
6707 * pointer to the next one after it. The last entry in the
6708 * the table is a NULL entry, so when we get to it, set cookie
6709 * to NULL and return that, indicating end of list, erstwhile
6710 * if we come in with cookie set to NULL, we are starting anew
6711 * at the front of the list.
6712 */
6713 if (cookie != NULL) {
6714 ta = fr_findtunebycookie(cookie, &tu.ipft_cookie, ifs);
6715 } else {
6716 ta = ifs->ifs_ipf_tuneables;
6717 tu.ipft_cookie = ta + 1;
6718 }
6719 if (ta != NULL) {
6720 /*
6721 * Entry found, but does the data pointed to by that
6722 * row fit in what we can return?
6723 */
6724 if (ta->ipft_sz > sizeof(tu.ipft_un))
6725 return EINVAL;
6726
6727 tu.ipft_vlong = 0;
6728 if (ta->ipft_sz == sizeof(u_long))
6729 tu.ipft_vlong = *ta->ipft_plong;
6730 else if (ta->ipft_sz == sizeof(u_int))
6731 tu.ipft_vint = *ta->ipft_pint;
6732 else if (ta->ipft_sz == sizeof(u_short))
6733 tu.ipft_vshort = *ta->ipft_pshort;
6734 else if (ta->ipft_sz == sizeof(u_char))
6735 tu.ipft_vchar = *ta->ipft_pchar;
6736
6737 tu.ipft_sz = ta->ipft_sz;
6738 tu.ipft_min = ta->ipft_min;
6739 tu.ipft_max = ta->ipft_max;
6740 tu.ipft_flags = ta->ipft_flags;
6741 bcopy(ta->ipft_name, tu.ipft_name,
6742 MIN(sizeof(tu.ipft_name),
6743 strlen(ta->ipft_name) + 1));
6744 }
6745 error = fr_outobj(data, &tu, IPFOBJ_TUNEABLE);
6746 break;
6747
6748 case SIOCIPFGET :
6749 case SIOCIPFSET :
6750 /*
6751 * Search by name or by cookie value for a particular entry
6752 * in the tuning paramter table.
6753 */
6754 error = ESRCH;
6755 if (cookie != NULL) {
6756 ta = fr_findtunebycookie(cookie, NULL, ifs);
6757 if (ta != NULL)
6758 error = 0;
6759 } else if (tu.ipft_name[0] != '\0') {
6760 ta = fr_findtunebyname(tu.ipft_name, ifs);
6761 if (ta != NULL)
6762 error = 0;
6763 }
6764 if (error != 0)
6765 break;
6766
6767 if (cmd == (ioctlcmd_t)SIOCIPFGET) {
6768 /*
6769 * Fetch the tuning parameters for a particular value
6770 */
6771 tu.ipft_vlong = 0;
6772 if (ta->ipft_sz == sizeof(u_long))
6773 tu.ipft_vlong = *ta->ipft_plong;
6774 else if (ta->ipft_sz == sizeof(u_int))
6775 tu.ipft_vint = *ta->ipft_pint;
6776 else if (ta->ipft_sz == sizeof(u_short))
6777 tu.ipft_vshort = *ta->ipft_pshort;
6778 else if (ta->ipft_sz == sizeof(u_char))
6779 tu.ipft_vchar = *ta->ipft_pchar;
6780 tu.ipft_cookie = ta;
6781 tu.ipft_sz = ta->ipft_sz;
6782 tu.ipft_min = ta->ipft_min;
6783 tu.ipft_max = ta->ipft_max;
6784 tu.ipft_flags = ta->ipft_flags;
6785 error = fr_outobj(data, &tu, IPFOBJ_TUNEABLE);
6786
6787 } else if (cmd == (ioctlcmd_t)SIOCIPFSET) {
6788 /*
6789 * Set an internal parameter. The hard part here is
6790 * getting the new value safely and correctly out of
6791 * the kernel (given we only know its size, not type.)
6792 */
6793 u_long in;
6794
6795 if (((ta->ipft_flags & IPFT_WRDISABLED) != 0) &&
6796 (ifs->ifs_fr_running > 0)) {
6797 error = EBUSY;
6798 break;
6799 }
6800
6801 in = tu.ipft_vlong;
6802 if (in < ta->ipft_min || in > ta->ipft_max) {
6803 error = EINVAL;
6804 break;
6805 }
6806
6807 if (ta->ipft_sz == sizeof(u_long)) {
6808 tu.ipft_vlong = *ta->ipft_plong;
6809 *ta->ipft_plong = in;
6810 } else if (ta->ipft_sz == sizeof(u_int)) {
6811 tu.ipft_vint = *ta->ipft_pint;
6812 *ta->ipft_pint = (u_int)(in & 0xffffffff);
6813 } else if (ta->ipft_sz == sizeof(u_short)) {
6814 tu.ipft_vshort = *ta->ipft_pshort;
6815 *ta->ipft_pshort = (u_short)(in & 0xffff);
6816 } else if (ta->ipft_sz == sizeof(u_char)) {
6817 tu.ipft_vchar = *ta->ipft_pchar;
6818 *ta->ipft_pchar = (u_char)(in & 0xff);
6819 }
6820 error = fr_outobj(data, &tu, IPFOBJ_TUNEABLE);
6821 }
6822 break;
6823
6824 default :
6825 error = EINVAL;
6826 break;
6827 }
6828
6829 return error;
6830 }
6831
6832
6833 /* ------------------------------------------------------------------------ */
6834 /* Function: fr_initialise */
6835 /* Returns: int - 0 == success, < 0 == failure */
6836 /* Parameters: None. */
6837 /* */
6838 /* Call of the initialise functions for all the various subsystems inside */
6839 /* of IPFilter. If any of them should fail, return immeadiately a failure */
6840 /* BUT do not try to recover from the error here. */
6841 /* ------------------------------------------------------------------------ */
fr_initialise(ifs)6842 int fr_initialise(ifs)
6843 ipf_stack_t *ifs;
6844 {
6845 int i;
6846
6847 #ifdef IPFILTER_LOG
6848 i = fr_loginit(ifs);
6849 if (i < 0)
6850 return -10 + i;
6851 #endif
6852 i = fr_natinit(ifs);
6853 if (i < 0)
6854 return -20 + i;
6855
6856 i = fr_stateinit(ifs);
6857 if (i < 0)
6858 return -30 + i;
6859
6860 i = fr_authinit(ifs);
6861 if (i < 0)
6862 return -40 + i;
6863
6864 i = fr_fraginit(ifs);
6865 if (i < 0)
6866 return -50 + i;
6867
6868 i = appr_init(ifs);
6869 if (i < 0)
6870 return -60 + i;
6871
6872 #ifdef IPFILTER_SYNC
6873 i = ipfsync_init(ifs);
6874 if (i < 0)
6875 return -70 + i;
6876 #endif
6877 #ifdef IPFILTER_SCAN
6878 i = ipsc_init(ifs);
6879 if (i < 0)
6880 return -80 + i;
6881 #endif
6882 #ifdef IPFILTER_LOOKUP
6883 i = ip_lookup_init(ifs);
6884 if (i < 0)
6885 return -90 + i;
6886 #endif
6887 #ifdef IPFILTER_COMPILED
6888 ipfrule_add(ifs);
6889 #endif
6890 return 0;
6891 }
6892
6893
6894 /* ------------------------------------------------------------------------ */
6895 /* Function: fr_deinitialise */
6896 /* Returns: None. */
6897 /* Parameters: None. */
6898 /* */
6899 /* Call all the various subsystem cleanup routines to deallocate memory or */
6900 /* destroy locks or whatever they've done that they need to now undo. */
6901 /* The order here IS important as there are some cross references of */
6902 /* internal data structures. */
6903 /* ------------------------------------------------------------------------ */
fr_deinitialise(ifs)6904 void fr_deinitialise(ifs)
6905 ipf_stack_t *ifs;
6906 {
6907 fr_fragunload(ifs);
6908 fr_authunload(ifs);
6909 fr_natunload(ifs);
6910 fr_stateunload(ifs);
6911 #ifdef IPFILTER_SCAN
6912 fr_scanunload(ifs);
6913 #endif
6914 appr_unload(ifs);
6915
6916 #ifdef IPFILTER_COMPILED
6917 ipfrule_remove(ifs);
6918 #endif
6919
6920 (void) frflush(IPL_LOGIPF, 0, FR_INQUE|FR_OUTQUE|FR_INACTIVE, ifs);
6921 (void) frflush(IPL_LOGIPF, 0, FR_INQUE|FR_OUTQUE, ifs);
6922 (void) frflush(IPL_LOGCOUNT, 0, FR_INQUE|FR_OUTQUE|FR_INACTIVE, ifs);
6923 (void) frflush(IPL_LOGCOUNT, 0, FR_INQUE|FR_OUTQUE, ifs);
6924
6925 #ifdef IPFILTER_LOOKUP
6926 ip_lookup_unload(ifs);
6927 #endif
6928
6929 #ifdef IPFILTER_LOG
6930 fr_logunload(ifs);
6931 #endif
6932 }
6933
6934
6935 /* ------------------------------------------------------------------------ */
6936 /* Function: fr_zerostats */
6937 /* Returns: int - 0 = success, else failure */
6938 /* Parameters: data(O) - pointer to pointer for copying data back to */
6939 /* */
6940 /* Copies the current statistics out to userspace and then zero's the */
6941 /* current ones in the kernel. The lock is only held across the bzero() as */
6942 /* the copyout may result in paging (ie network activity.) */
6943 /* ------------------------------------------------------------------------ */
fr_zerostats(data,ifs)6944 int fr_zerostats(data, ifs)
6945 caddr_t data;
6946 ipf_stack_t *ifs;
6947 {
6948 friostat_t fio;
6949 int error;
6950
6951 fr_getstat(&fio, ifs);
6952 error = copyoutptr(&fio, data, sizeof(fio));
6953 if (error)
6954 return EFAULT;
6955
6956 WRITE_ENTER(&ifs->ifs_ipf_mutex);
6957 bzero((char *)ifs->ifs_frstats, sizeof(*ifs->ifs_frstats) * 2);
6958 RWLOCK_EXIT(&ifs->ifs_ipf_mutex);
6959
6960 return 0;
6961 }
6962
6963
6964 #ifdef _KERNEL
6965 /* ------------------------------------------------------------------------ */
6966 /* Function: fr_resolvedest */
6967 /* Returns: Nil */
6968 /* Parameters: fdp(IO) - pointer to destination information to resolve */
6969 /* v(I) - IP protocol version to match */
6970 /* */
6971 /* Looks up an interface name in the frdest structure pointed to by fdp and */
6972 /* if a matching name can be found for the particular IP protocol version */
6973 /* then store the interface pointer in the frdest struct. If no match is */
6974 /* found, then set the interface pointer to be -1 as NULL is considered to */
6975 /* indicate there is no information at all in the structure. */
6976 /* ------------------------------------------------------------------------ */
fr_resolvedest(fdp,v,ifs)6977 void fr_resolvedest(fdp, v, ifs)
6978 frdest_t *fdp;
6979 int v;
6980 ipf_stack_t *ifs;
6981 {
6982 fdp->fd_ifp = NULL;
6983
6984 if (*fdp->fd_ifname != '\0') {
6985 fdp->fd_ifp = GETIFP(fdp->fd_ifname, v, ifs);
6986 if (fdp->fd_ifp == NULL)
6987 fdp->fd_ifp = (void *)-1;
6988 }
6989 }
6990 #endif /* _KERNEL */
6991
6992
6993 /* ------------------------------------------------------------------------ */
6994 /* Function: fr_resolvenic */
6995 /* Returns: void* - NULL = wildcard name, -1 = failed to find NIC, else */
6996 /* pointer to interface structure for NIC */
6997 /* Parameters: name(I) - complete interface name */
6998 /* v(I) - IP protocol version */
6999 /* */
7000 /* Look for a network interface structure that firstly has a matching name */
7001 /* to that passed in and that is also being used for that IP protocol */
7002 /* version (necessary on some platforms where there are separate listings */
7003 /* for both IPv4 and IPv6 on the same physical NIC. */
7004 /* */
7005 /* One might wonder why name gets terminated with a \0 byte in here. The */
7006 /* reason is an interface name could get into the kernel structures of ipf */
7007 /* in any number of ways and so long as they all use the same sized array */
7008 /* to put the name in, it makes sense to ensure it gets null terminated */
7009 /* before it is used for its intended purpose - finding its match in the */
7010 /* kernel's list of configured interfaces. */
7011 /* */
7012 /* NOTE: This SHOULD ONLY be used with IPFilter structures that have an */
7013 /* array for the name that is LIFNAMSIZ bytes (at least) in length. */
7014 /* ------------------------------------------------------------------------ */
fr_resolvenic(name,v,ifs)7015 void *fr_resolvenic(name, v, ifs)
7016 char *name;
7017 int v;
7018 ipf_stack_t *ifs;
7019 {
7020 void *nic;
7021
7022 if (name[0] == '\0')
7023 return NULL;
7024
7025 if ((name[1] == '\0') && ((name[0] == '-') || (name[0] == '*'))) {
7026 return NULL;
7027 }
7028
7029 name[LIFNAMSIZ - 1] = '\0';
7030
7031 nic = GETIFP(name, v, ifs);
7032 if (nic == NULL)
7033 nic = (void *)-1;
7034 return nic;
7035 }
7036
7037
7038 /* ------------------------------------------------------------------------ */
7039 /* Function: ipf_expiretokens */
7040 /* Returns: None. */
7041 /* Parameters: ifs - ipf stack instance */
7042 /* */
7043 /* This function is run every ipf tick to see if there are any tokens that */
7044 /* have been held for too long and need to be freed up. */
7045 /* ------------------------------------------------------------------------ */
ipf_expiretokens(ifs)7046 void ipf_expiretokens(ifs)
7047 ipf_stack_t *ifs;
7048 {
7049 ipftoken_t *it;
7050
7051 WRITE_ENTER(&ifs->ifs_ipf_tokens);
7052 while ((it = ifs->ifs_ipftokenhead) != NULL) {
7053 if (it->ipt_die > ifs->ifs_fr_ticks)
7054 break;
7055
7056 ipf_freetoken(it, ifs);
7057 }
7058 RWLOCK_EXIT(&ifs->ifs_ipf_tokens);
7059 }
7060
7061
7062 /* ------------------------------------------------------------------------ */
7063 /* Function: ipf_deltoken */
7064 /* Returns: int - 0 = success, else error */
7065 /* Parameters: type(I) - the token type to match */
7066 /* uid(I) - uid owning the token */
7067 /* ptr(I) - context pointer for the token */
7068 /* ifs - ipf stack instance */
7069 /* */
7070 /* This function looks for a a token in the current list that matches up */
7071 /* the fields (type, uid, ptr). If none is found, ESRCH is returned, else */
7072 /* call ipf_freetoken() to remove it from the list. */
7073 /* ------------------------------------------------------------------------ */
ipf_deltoken(type,uid,ptr,ifs)7074 int ipf_deltoken(type, uid, ptr, ifs)
7075 int type, uid;
7076 void *ptr;
7077 ipf_stack_t *ifs;
7078 {
7079 ipftoken_t *it;
7080 int error = ESRCH;
7081
7082 WRITE_ENTER(&ifs->ifs_ipf_tokens);
7083 for (it = ifs->ifs_ipftokenhead; it != NULL; it = it->ipt_next)
7084 if (ptr == it->ipt_ctx && type == it->ipt_type &&
7085 uid == it->ipt_uid) {
7086 ipf_freetoken(it, ifs);
7087 error = 0;
7088 break;
7089 }
7090 RWLOCK_EXIT(&ifs->ifs_ipf_tokens);
7091
7092 return error;
7093 }
7094
7095
7096 /* ------------------------------------------------------------------------ */
7097 /* Function: ipf_unlinktoken */
7098 /* Returns: None. */
7099 /* Parameters: token(I) - pointer to token structure */
7100 /* ifs - ipf stack instance */
7101 /* */
7102 /* This function unlinks a token structure from the linked list of tokens */
7103 /* that it belongs to. The head pointer never needs to be explicitly */
7104 /* adjusted, but the tail does due to the linked list implementation. */
7105 /* ------------------------------------------------------------------------ */
ipf_unlinktoken(token,ifs)7106 static void ipf_unlinktoken(token, ifs)
7107 ipftoken_t *token;
7108 ipf_stack_t *ifs;
7109 {
7110
7111 if (ifs->ifs_ipftokentail == &token->ipt_next)
7112 ifs->ifs_ipftokentail = token->ipt_pnext;
7113
7114 *token->ipt_pnext = token->ipt_next;
7115 if (token->ipt_next != NULL)
7116 token->ipt_next->ipt_pnext = token->ipt_pnext;
7117 }
7118
7119
7120 /* ------------------------------------------------------------------------ */
7121 /* Function: ipf_findtoken */
7122 /* Returns: ipftoken_t * - NULL if no memory, else pointer to token */
7123 /* Parameters: type(I) - the token type to match */
7124 /* uid(I) - uid owning the token */
7125 /* ptr(I) - context pointer for the token */
7126 /* ifs - ipf stack instance */
7127 /* */
7128 /* This function looks for a live token in the list of current tokens that */
7129 /* matches the tuple (type, uid, ptr). If one cannot be found then one is */
7130 /* allocated. If one is found then it is moved to the top of the list of */
7131 /* currently active tokens. */
7132 /* */
7133 /* NOTE: It is by design that this function returns holding a read lock on */
7134 /* ipf_tokens. Callers must make sure they release it! */
7135 /* ------------------------------------------------------------------------ */
ipf_findtoken(type,uid,ptr,ifs)7136 ipftoken_t *ipf_findtoken(type, uid, ptr, ifs)
7137 int type, uid;
7138 void *ptr;
7139 ipf_stack_t *ifs;
7140 {
7141 ipftoken_t *it, *new;
7142
7143 KMALLOC(new, ipftoken_t *);
7144
7145 WRITE_ENTER(&ifs->ifs_ipf_tokens);
7146 for (it = ifs->ifs_ipftokenhead; it != NULL; it = it->ipt_next) {
7147 if (it->ipt_alive == 0)
7148 continue;
7149 if (ptr == it->ipt_ctx && type == it->ipt_type &&
7150 uid == it->ipt_uid)
7151 break;
7152 }
7153
7154 if (it == NULL) {
7155 it = new;
7156 new = NULL;
7157 if (it == NULL)
7158 return NULL;
7159 it->ipt_data = NULL;
7160 it->ipt_ctx = ptr;
7161 it->ipt_uid = uid;
7162 it->ipt_type = type;
7163 it->ipt_next = NULL;
7164 it->ipt_alive = 1;
7165 } else {
7166 if (new != NULL) {
7167 KFREE(new);
7168 new = NULL;
7169 }
7170
7171 ipf_unlinktoken(it, ifs);
7172 }
7173 it->ipt_pnext = ifs->ifs_ipftokentail;
7174 *ifs->ifs_ipftokentail = it;
7175 ifs->ifs_ipftokentail = &it->ipt_next;
7176 it->ipt_next = NULL;
7177
7178 it->ipt_die = ifs->ifs_fr_ticks + 2;
7179
7180 MUTEX_DOWNGRADE(&ifs->ifs_ipf_tokens);
7181
7182 return it;
7183 }
7184
7185
7186 /* ------------------------------------------------------------------------ */
7187 /* Function: ipf_freetoken */
7188 /* Returns: None. */
7189 /* Parameters: token(I) - pointer to token structure */
7190 /* ifs - ipf stack instance */
7191 /* */
7192 /* This function unlinks a token from the linked list and on the path to */
7193 /* free'ing the data, it calls the dereference function that is associated */
7194 /* with the type of data pointed to by the token as it is considered to */
7195 /* hold a reference to it. */
7196 /* ------------------------------------------------------------------------ */
ipf_freetoken(token,ifs)7197 void ipf_freetoken(token, ifs)
7198 ipftoken_t *token;
7199 ipf_stack_t *ifs;
7200 {
7201 void *data, **datap;
7202
7203 ipf_unlinktoken(token, ifs);
7204
7205 data = token->ipt_data;
7206 datap = &data;
7207
7208 if ((data != NULL) && (data != (void *)-1)) {
7209 switch (token->ipt_type)
7210 {
7211 case IPFGENITER_IPF :
7212 (void)fr_derefrule((frentry_t **)datap, ifs);
7213 break;
7214 case IPFGENITER_IPNAT :
7215 WRITE_ENTER(&ifs->ifs_ipf_nat);
7216 fr_ipnatderef((ipnat_t **)datap, ifs);
7217 RWLOCK_EXIT(&ifs->ifs_ipf_nat);
7218 break;
7219 case IPFGENITER_NAT :
7220 fr_natderef((nat_t **)datap, ifs);
7221 break;
7222 case IPFGENITER_STATE :
7223 fr_statederef((ipstate_t **)datap, ifs);
7224 break;
7225 case IPFGENITER_FRAG :
7226 fr_fragderef((ipfr_t **)datap, &ifs->ifs_ipf_frag, ifs);
7227 break;
7228 case IPFGENITER_NATFRAG :
7229 fr_fragderef((ipfr_t **)datap,
7230 &ifs->ifs_ipf_natfrag, ifs);
7231 break;
7232 case IPFGENITER_HOSTMAP :
7233 WRITE_ENTER(&ifs->ifs_ipf_nat);
7234 fr_hostmapdel((hostmap_t **)datap);
7235 RWLOCK_EXIT(&ifs->ifs_ipf_nat);
7236 break;
7237 default :
7238 (void) ip_lookup_iterderef(token->ipt_type, data, ifs);
7239 break;
7240 }
7241 }
7242
7243 KFREE(token);
7244 }
7245
7246
7247 /* ------------------------------------------------------------------------ */
7248 /* Function: ipf_getnextrule */
7249 /* Returns: int - 0 = success, else error */
7250 /* Parameters: t(I) - pointer to destination information to resolve */
7251 /* ptr(I) - pointer to ipfobj_t to copyin from user space */
7252 /* ifs - ipf stack instance */
7253 /* */
7254 /* This function's first job is to bring in the ipfruleiter_t structure via */
7255 /* the ipfobj_t structure to determine what should be the next rule to */
7256 /* return. Once the ipfruleiter_t has been brought in, it then tries to */
7257 /* find the 'next rule'. This may include searching rule group lists or */
7258 /* just be as simple as looking at the 'next' field in the rule structure. */
7259 /* When we have found the rule to return, increase its reference count and */
7260 /* if we used an existing rule to get here, decrease its reference count. */
7261 /* ------------------------------------------------------------------------ */
ipf_getnextrule(t,ptr,ifs)7262 int ipf_getnextrule(t, ptr, ifs)
7263 ipftoken_t *t;
7264 void *ptr;
7265 ipf_stack_t *ifs;
7266 {
7267 frentry_t *fr, *next, zero;
7268 int error, out, count;
7269 ipfruleiter_t it;
7270 frgroup_t *fg;
7271 char *dst;
7272
7273 if (t == NULL || ptr == NULL)
7274 return EFAULT;
7275 error = fr_inobj(ptr, &it, IPFOBJ_IPFITER);
7276 if (error != 0)
7277 return error;
7278 if ((it.iri_ver != AF_INET) && (it.iri_ver != AF_INET6))
7279 return EINVAL;
7280 if ((it.iri_inout < 0) || (it.iri_inout > 3))
7281 return EINVAL;
7282 if (it.iri_nrules == 0)
7283 return EINVAL;
7284 if ((it.iri_active != 0) && (it.iri_active != 1))
7285 return EINVAL;
7286 if (it.iri_rule == NULL)
7287 return EFAULT;
7288
7289 /*
7290 * Use bitmask on it.iri_inout to determine direction.
7291 * F_OUT (1) and F_ACOUT (3) mask to out = 1, while
7292 * F_IN (0) and F_ACIN (2) mask to out = 0.
7293 */
7294 out = it.iri_inout & F_OUT;
7295 READ_ENTER(&ifs->ifs_ipf_mutex);
7296
7297 /*
7298 * Retrieve "previous" entry from token and find the next entry.
7299 */
7300 fr = t->ipt_data;
7301 if (fr == NULL) {
7302 if (*it.iri_group == '\0') {
7303 /*
7304 * Use bitmask again to determine accounting or not.
7305 * F_ACIN will mask to accounting cases F_ACIN (2)
7306 * or F_ACOUT (3), but not F_IN or F_OUT.
7307 */
7308 if ((it.iri_inout & F_ACIN) != 0) {
7309 if (it.iri_ver == AF_INET)
7310 next = ifs->ifs_ipacct
7311 [out][it.iri_active];
7312 else
7313 next = ifs->ifs_ipacct6
7314 [out][it.iri_active];
7315 } else {
7316 if (it.iri_ver == AF_INET)
7317 next = ifs->ifs_ipfilter
7318 [out][it.iri_active];
7319 else
7320 next = ifs->ifs_ipfilter6
7321 [out][it.iri_active];
7322 }
7323 } else {
7324 fg = fr_findgroup(it.iri_group, IPL_LOGIPF,
7325 it.iri_active, NULL, ifs);
7326 if (fg != NULL)
7327 next = fg->fg_start;
7328 else
7329 next = NULL;
7330 }
7331 } else {
7332 next = fr->fr_next;
7333 }
7334
7335 dst = (char *)it.iri_rule;
7336 /*
7337 * The ipfruleiter may ask for more than 1 rule at a time to be
7338 * copied out, so long as that many exist in the list to start with!
7339 */
7340 for (count = it.iri_nrules; count > 0; count--) {
7341 /*
7342 * If we found an entry, add reference to it and update token.
7343 * Otherwise, zero out data to be returned and NULL out token.
7344 */
7345 if (next != NULL) {
7346 MUTEX_ENTER(&next->fr_lock);
7347 next->fr_ref++;
7348 MUTEX_EXIT(&next->fr_lock);
7349 t->ipt_data = next;
7350 } else {
7351 bzero(&zero, sizeof(zero));
7352 next = &zero;
7353 t->ipt_data = NULL;
7354 }
7355
7356 /*
7357 * Now that we have ref, it's save to give up lock.
7358 */
7359 RWLOCK_EXIT(&ifs->ifs_ipf_mutex);
7360
7361 /*
7362 * Copy out data and clean up references and token as needed.
7363 */
7364 error = COPYOUT(next, dst, sizeof(*next));
7365 if (error != 0)
7366 error = EFAULT;
7367 if (t->ipt_data == NULL) {
7368 ipf_freetoken(t, ifs);
7369 break;
7370 } else {
7371 if (fr != NULL)
7372 (void) fr_derefrule(&fr, ifs);
7373 if (next->fr_data != NULL) {
7374 dst += sizeof(*next);
7375 error = COPYOUT(next->fr_data, dst,
7376 next->fr_dsize);
7377 if (error != 0)
7378 error = EFAULT;
7379 else
7380 dst += next->fr_dsize;
7381 }
7382 if (next->fr_next == NULL) {
7383 ipf_freetoken(t, ifs);
7384 break;
7385 }
7386 }
7387
7388 if ((count == 1) || (error != 0))
7389 break;
7390
7391 READ_ENTER(&ifs->ifs_ipf_mutex);
7392 fr = next;
7393 next = fr->fr_next;
7394 }
7395
7396 return error;
7397 }
7398
7399
7400 /* ------------------------------------------------------------------------ */
7401 /* Function: fr_frruleiter */
7402 /* Returns: int - 0 = success, else error */
7403 /* Parameters: data(I) - the token type to match */
7404 /* uid(I) - uid owning the token */
7405 /* ptr(I) - context pointer for the token */
7406 /* ifs - ipf stack instance */
7407 /* */
7408 /* This function serves as a stepping stone between fr_ipf_ioctl and */
7409 /* ipf_getnextrule. It's role is to find the right token in the kernel for */
7410 /* the process doing the ioctl and use that to ask for the next rule. */
7411 /* ------------------------------------------------------------------------ */
ipf_frruleiter(data,uid,ctx,ifs)7412 int ipf_frruleiter(data, uid, ctx, ifs)
7413 void *data, *ctx;
7414 int uid;
7415 ipf_stack_t *ifs;
7416 {
7417 ipftoken_t *token;
7418 int error;
7419
7420 token = ipf_findtoken(IPFGENITER_IPF, uid, ctx, ifs);
7421 if (token != NULL)
7422 error = ipf_getnextrule(token, data, ifs);
7423 else
7424 error = EFAULT;
7425 RWLOCK_EXIT(&ifs->ifs_ipf_tokens);
7426
7427 return error;
7428 }
7429
7430
7431 /* ------------------------------------------------------------------------ */
7432 /* Function: ipf_geniter */
7433 /* Returns: int - 0 = success, else error */
7434 /* Parameters: token(I) - pointer to ipftoken structure */
7435 /* itp(I) - pointer to ipfgeniter structure */
7436 /* ifs - ipf stack instance */
7437 /* */
7438 /* Generic iterator called from ipf_genericiter. Currently only used for */
7439 /* walking through list of fragments. */
7440 /* ------------------------------------------------------------------------ */
ipf_geniter(token,itp,ifs)7441 int ipf_geniter(token, itp, ifs)
7442 ipftoken_t *token;
7443 ipfgeniter_t *itp;
7444 ipf_stack_t *ifs;
7445 {
7446 int error;
7447
7448 switch (itp->igi_type)
7449 {
7450 case IPFGENITER_FRAG :
7451 error = fr_nextfrag(token, itp, &ifs->ifs_ipfr_list,
7452 &ifs->ifs_ipfr_tail, &ifs->ifs_ipf_frag,
7453 ifs);
7454 break;
7455 default :
7456 error = EINVAL;
7457 break;
7458 }
7459
7460 return error;
7461 }
7462
7463
7464 /* ------------------------------------------------------------------------ */
7465 /* Function: ipf_genericiter */
7466 /* Returns: int - 0 = success, else error */
7467 /* Parameters: data(I) - the token type to match */
7468 /* uid(I) - uid owning the token */
7469 /* ptr(I) - context pointer for the token */
7470 /* ifs - ipf stack instance */
7471 /* */
7472 /* This function serves as a stepping stone between fr_ipf_ioctl and */
7473 /* ipf_geniter when handling SIOCGENITER. It's role is to find the right */
7474 /* token in the kernel for the process using the ioctl, and to use that */
7475 /* token when calling ipf_geniter. */
7476 /* ------------------------------------------------------------------------ */
ipf_genericiter(data,uid,ctx,ifs)7477 int ipf_genericiter(data, uid, ctx, ifs)
7478 void *data, *ctx;
7479 int uid;
7480 ipf_stack_t *ifs;
7481 {
7482 ipftoken_t *token;
7483 ipfgeniter_t iter;
7484 int error;
7485
7486 error = fr_inobj(data, &iter, IPFOBJ_GENITER);
7487 if (error != 0)
7488 return error;
7489
7490 token = ipf_findtoken(iter.igi_type, uid, ctx, ifs);
7491 if (token != NULL) {
7492 token->ipt_subtype = iter.igi_type;
7493 error = ipf_geniter(token, &iter, ifs);
7494 } else
7495 error = EFAULT;
7496 RWLOCK_EXIT(&ifs->ifs_ipf_tokens);
7497
7498 return error;
7499 }
7500
7501
7502 /* --------------------------------------------------------------------- */
7503 /* Function: ipf_earlydrop */
7504 /* Returns: number of dropped/removed entries from the queue */
7505 /* Parameters: flushtype - which table we're cleaning (NAT or State) */
7506 /* ifq - pointer to queue with entries to be deleted */
7507 /* idletime - entry must be idle this long to be deleted */
7508 /* ifs - ipf stack instance */
7509 /* */
7510 /* Function is invoked from state/NAT flush routines to remove entries */
7511 /* from specified timeout queue, based on how long they've sat idle, */
7512 /* without waiting for it to happen on its own. */
7513 /* --------------------------------------------------------------------- */
ipf_earlydrop(flushtype,ifq,idletime,ifs)7514 int ipf_earlydrop(flushtype, ifq, idletime, ifs)
7515 int flushtype;
7516 ipftq_t *ifq;
7517 int idletime;
7518 ipf_stack_t *ifs;
7519 {
7520 ipftqent_t *tqe, *tqn;
7521 unsigned int dropped;
7522 int droptick;
7523 void *ent;
7524
7525 if (ifq == NULL)
7526 return (0);
7527
7528 dropped = 0;
7529
7530 /*
7531 * Determine the tick representing the idle time we're interested
7532 * in. If an entry exists in the queue, and it was touched before
7533 * that tick, then it's been idle longer than idletime, so it should
7534 * be deleted.
7535 */
7536 droptick = ifs->ifs_fr_ticks - idletime;
7537 tqn = ifq->ifq_head;
7538 while ((tqe = tqn) != NULL && tqe->tqe_touched < droptick) {
7539 tqn = tqe->tqe_next;
7540 ent = tqe->tqe_parent;
7541 switch (flushtype)
7542 {
7543 case NAT_FLUSH:
7544 if (nat_delete((nat_t *)ent, NL_FLUSH, ifs) == 0)
7545 dropped++;
7546 break;
7547 case STATE_FLUSH:
7548 if (fr_delstate((ipstate_t *)ent, ISL_FLUSH, ifs) == 0)
7549 dropped++;
7550 break;
7551 default:
7552 return (0);
7553 }
7554 }
7555 return (dropped);
7556 }
7557
7558
7559 /* --------------------------------------------------------------------- */
7560 /* Function: ipf_flushclosing */
7561 /* Returns: int - number of entries deleted */
7562 /* Parameters: flushtype - which table we're cleaning (NAT or State) */
7563 /* stateval - TCP state at which to start removing entries */
7564 /* ipfqs - pointer to timeout queues */
7565 /* userqs - pointer to user defined queues */
7566 /* ifs - ipf stack instance */
7567 /* */
7568 /* Remove state/NAT table entries for TCP connections which are in the */
7569 /* process of closing, and have at least reached the state specified by */
7570 /* the 'stateval' parameter. */
7571 /* --------------------------------------------------------------------- */
ipf_flushclosing(flushtype,stateval,ipfqs,userqs,ifs)7572 int ipf_flushclosing(flushtype, stateval, ipfqs, userqs, ifs)
7573 int flushtype, stateval;
7574 ipftq_t *ipfqs, *userqs;
7575 ipf_stack_t *ifs;
7576 {
7577 ipftq_t *ifq, *ifqn;
7578 ipftqent_t *tqe, *tqn;
7579 int dropped;
7580 void *ent;
7581 nat_t *nat;
7582 ipstate_t *is;
7583
7584 dropped = 0;
7585
7586 /*
7587 * Start by deleting any entries in specific timeout queues.
7588 */
7589 ifqn = &ipfqs[stateval];
7590 while ((ifq = ifqn) != NULL) {
7591 ifqn = ifq->ifq_next;
7592 dropped += ipf_earlydrop(flushtype, ifq, (int)0, ifs);
7593 }
7594
7595 /*
7596 * Next, look through user defined queues for closing entries.
7597 */
7598 ifqn = userqs;
7599 while ((ifq = ifqn) != NULL) {
7600 ifqn = ifq->ifq_next;
7601 tqn = ifq->ifq_head;
7602 while ((tqe = tqn) != NULL) {
7603 tqn = tqe->tqe_next;
7604 ent = tqe->tqe_parent;
7605 switch (flushtype)
7606 {
7607 case NAT_FLUSH:
7608 nat = (nat_t *)ent;
7609 if ((nat->nat_p == IPPROTO_TCP) &&
7610 (nat->nat_tcpstate[0] >= stateval) &&
7611 (nat->nat_tcpstate[1] >= stateval) &&
7612 (nat_delete(nat, NL_EXPIRE, ifs) == 0))
7613 dropped++;
7614 break;
7615 case STATE_FLUSH:
7616 is = (ipstate_t *)ent;
7617 if ((is->is_p == IPPROTO_TCP) &&
7618 (is->is_state[0] >= stateval) &&
7619 (is->is_state[1] >= stateval) &&
7620 (fr_delstate(is, ISL_EXPIRE, ifs) == 0))
7621 dropped++;
7622 break;
7623 default:
7624 return (0);
7625 }
7626 }
7627 }
7628 return (dropped);
7629 }
7630
7631
7632 /* --------------------------------------------------------------------- */
7633 /* Function: ipf_extraflush */
7634 /* Returns: int - number of entries flushed (0 = none) */
7635 /* Parameters: flushtype - which table we're cleaning (NAT or State) */
7636 /* ipfqs - pointer to 'established' timeout queue */
7637 /* userqs - pointer to user defined queues */
7638 /* ifs - ipf stack instance */
7639 /* */
7640 /* This function gets called when either NAT or state tables fill up. */
7641 /* We need to try a bit harder to free up some space. The function will */
7642 /* flush entries for TCP connections which have been idle a long time. */
7643 /* */
7644 /* Currently, the idle time is checked using values from ideltime_tab[] */
7645 /* --------------------------------------------------------------------- */
ipf_extraflush(flushtype,ipfqs,userqs,ifs)7646 int ipf_extraflush(flushtype, ipfqs, userqs, ifs)
7647 int flushtype;
7648 ipftq_t *ipfqs, *userqs;
7649 ipf_stack_t *ifs;
7650 {
7651 ipftq_t *ifq, *ifqn;
7652 int idletime, removed, idle_idx;
7653
7654 removed = 0;
7655
7656 /*
7657 * Determine initial threshold for minimum idle time based on
7658 * how long ipfilter has been running. Ipfilter needs to have
7659 * been up as long as the smallest interval to continue on.
7660 *
7661 * Minimum idle times stored in idletime_tab and indexed by
7662 * idle_idx. Start at upper end of array and work backwards.
7663 *
7664 * Once the index is found, set the initial idle time to the
7665 * first interval before the current ipfilter run time.
7666 */
7667 if (ifs->ifs_fr_ticks < idletime_tab[0])
7668 return (0);
7669 idle_idx = (sizeof (idletime_tab) / sizeof (int)) - 1;
7670 if (ifs->ifs_fr_ticks > idletime_tab[idle_idx]) {
7671 idletime = idletime_tab[idle_idx];
7672 } else {
7673 while ((idle_idx > 0) &&
7674 (ifs->ifs_fr_ticks < idletime_tab[idle_idx]))
7675 idle_idx--;
7676
7677 idletime = (ifs->ifs_fr_ticks /
7678 idletime_tab[idle_idx]) *
7679 idletime_tab[idle_idx];
7680 }
7681
7682 while (idle_idx >= 0) {
7683 /*
7684 * Check to see if we need to delete more entries.
7685 * If we do, start with appropriate timeout queue.
7686 */
7687 if (flushtype == NAT_FLUSH) {
7688 if (NAT_TAB_WATER_LEVEL(ifs) <=
7689 ifs->ifs_nat_flush_level_lo)
7690 break;
7691 } else if (flushtype == STATE_FLUSH) {
7692 if (ST_TAB_WATER_LEVEL(ifs) <=
7693 ifs->ifs_state_flush_level_lo)
7694 break;
7695 } else {
7696 break;
7697 }
7698
7699 removed += ipf_earlydrop(flushtype, ipfqs, idletime, ifs);
7700
7701 /*
7702 * Next, check the user defined queues. But first, make
7703 * certain that timeout queue deletions didn't do enough.
7704 */
7705 if (flushtype == NAT_FLUSH) {
7706 if (NAT_TAB_WATER_LEVEL(ifs) <=
7707 ifs->ifs_nat_flush_level_lo)
7708 break;
7709 } else {
7710 if (ST_TAB_WATER_LEVEL(ifs) <=
7711 ifs->ifs_state_flush_level_lo)
7712 break;
7713 }
7714 ifqn = userqs;
7715 while ((ifq = ifqn) != NULL) {
7716 ifqn = ifq->ifq_next;
7717 removed += ipf_earlydrop(flushtype, ifq, idletime, ifs);
7718 }
7719
7720 /*
7721 * Adjust the granularity of idle time.
7722 *
7723 * If we reach an interval boundary, we need to
7724 * either adjust the idle time accordingly or exit
7725 * the loop altogether (if this is very last check).
7726 */
7727 idletime -= idletime_tab[idle_idx];
7728 if (idletime < idletime_tab[idle_idx]) {
7729 if (idle_idx != 0) {
7730 idletime = idletime_tab[idle_idx] -
7731 idletime_tab[idle_idx - 1];
7732 idle_idx--;
7733 } else {
7734 break;
7735 }
7736 }
7737 }
7738
7739 return (removed);
7740 }
7741