xref: /illumos-gate/usr/src/uts/common/inet/ipf/ip_fil_solaris.c (revision a6bde1a23b60f140c7ed78df979c2e22b1ed9b2c)
1 /*
2  * Copyright (C) 1993-2001, 2003 by Darren Reed.
3  *
4  * See the IPFILTER.LICENCE file for details on licencing.
5  *
6  * Copyright (c) 2003, 2010, Oracle and/or its affiliates. All rights reserved.
7  */
8 
9 #if !defined(lint)
10 static const char sccsid[] = "@(#)ip_fil_solaris.c	1.7 07/22/06 (C) 1993-2000 Darren Reed";
11 static const char rcsid[] = "@(#)$Id: ip_fil_solaris.c,v 2.62.2.19 2005/07/13 21:40:46 darrenr Exp $";
12 #endif
13 
14 #include <sys/types.h>
15 #include <sys/errno.h>
16 #include <sys/param.h>
17 #include <sys/cpuvar.h>
18 #include <sys/open.h>
19 #include <sys/ioctl.h>
20 #include <sys/filio.h>
21 #include <sys/systm.h>
22 #include <sys/strsubr.h>
23 #include <sys/cred.h>
24 #include <sys/ddi.h>
25 #include <sys/sunddi.h>
26 #include <sys/ksynch.h>
27 #include <sys/kmem.h>
28 #include <sys/mkdev.h>
29 #include <sys/protosw.h>
30 #include <sys/socket.h>
31 #include <sys/dditypes.h>
32 #include <sys/cmn_err.h>
33 #include <sys/zone.h>
34 #include <net/if.h>
35 #include <net/af.h>
36 #include <net/route.h>
37 #include <netinet/in.h>
38 #include <netinet/in_systm.h>
39 #include <netinet/ip.h>
40 #include <netinet/ip_var.h>
41 #include <netinet/tcp.h>
42 #include <netinet/udp.h>
43 #include <netinet/tcpip.h>
44 #include <netinet/ip_icmp.h>
45 #include "netinet/ip_compat.h"
46 #ifdef	USE_INET6
47 # include <netinet/icmp6.h>
48 #endif
49 #include "netinet/ip_fil.h"
50 #include "netinet/ip_nat.h"
51 #include "netinet/ip_frag.h"
52 #include "netinet/ip_state.h"
53 #include "netinet/ip_auth.h"
54 #include "netinet/ip_proxy.h"
55 #include "netinet/ipf_stack.h"
56 #ifdef	IPFILTER_LOOKUP
57 # include "netinet/ip_lookup.h"
58 #endif
59 #include <inet/ip_ire.h>
60 
61 #include <sys/md5.h>
62 #include <sys/neti.h>
63 
64 static	int	frzerostats __P((caddr_t, ipf_stack_t *));
65 static	int	fr_setipfloopback __P((int, ipf_stack_t *));
66 static	int	fr_enableipf __P((ipf_stack_t *, int));
67 static	int	fr_send_ip __P((fr_info_t *fin, mblk_t *m, mblk_t **mp));
68 static	int	ipf_nic_event_v4 __P((hook_event_token_t, hook_data_t, void *));
69 static	int	ipf_nic_event_v6 __P((hook_event_token_t, hook_data_t, void *));
70 static	int	ipf_hook __P((hook_data_t, int, int, void *));
71 static	int	ipf_hook4_in __P((hook_event_token_t, hook_data_t, void *));
72 static	int	ipf_hook4_out __P((hook_event_token_t, hook_data_t, void *));
73 static	int	ipf_hook4_loop_out __P((hook_event_token_t, hook_data_t,
74     void *));
75 static	int	ipf_hook4_loop_in __P((hook_event_token_t, hook_data_t, void *));
76 static	int	ipf_hook4 __P((hook_data_t, int, int, void *));
77 static	int	ipf_hook6_out __P((hook_event_token_t, hook_data_t, void *));
78 static	int	ipf_hook6_in __P((hook_event_token_t, hook_data_t, void *));
79 static	int	ipf_hook6_loop_out __P((hook_event_token_t, hook_data_t,
80     void *));
81 static	int	ipf_hook6_loop_in __P((hook_event_token_t, hook_data_t,
82     void *));
83 static	int     ipf_hook6 __P((hook_data_t, int, int, void *));
84 extern	int	ipf_geniter __P((ipftoken_t *, ipfgeniter_t *, ipf_stack_t *));
85 extern	int	ipf_frruleiter __P((void *, int, void *, ipf_stack_t *));
86 
87 #if SOLARIS2 < 10
88 #if SOLARIS2 >= 7
89 u_int		*ip_ttl_ptr = NULL;
90 u_int		*ip_mtudisc = NULL;
91 # if SOLARIS2 >= 8
92 int		*ip_forwarding = NULL;
93 u_int		*ip6_forwarding = NULL;
94 # else
95 u_int		*ip_forwarding = NULL;
96 # endif
97 #else
98 u_long		*ip_ttl_ptr = NULL;
99 u_long		*ip_mtudisc = NULL;
100 u_long		*ip_forwarding = NULL;
101 #endif
102 #endif
103 
104 
105 /* ------------------------------------------------------------------------ */
106 /* Function:    ipldetach                                                   */
107 /* Returns:     int - 0 == success, else error.                             */
108 /* Parameters:  Nil                                                         */
109 /*                                                                          */
110 /* This function is responsible for undoing anything that might have been   */
111 /* done in a call to iplattach().  It must be able to clean up from a call  */
112 /* to iplattach() that did not succeed.  Why might that happen?  Someone    */
113 /* configures a table to be so large that we cannot allocate enough memory  */
114 /* for it.                                                                  */
115 /* ------------------------------------------------------------------------ */
116 int ipldetach(ifs)
117 ipf_stack_t *ifs;
118 {
119 
120 	ASSERT(rw_read_locked(&ifs->ifs_ipf_global.ipf_lk) == 0);
121 
122 #if SOLARIS2 < 10
123 
124 	if (ifs->ifs_fr_control_forwarding & 2) {
125 		if (ip_forwarding != NULL)
126 			*ip_forwarding = 0;
127 #if SOLARIS2 >= 8
128 		if (ip6_forwarding != NULL)
129 			*ip6_forwarding = 0;
130 #endif
131 	}
132 #endif
133 
134 	/*
135 	 * This lock needs to be dropped around the net_hook_unregister calls
136 	 * because we can deadlock here with:
137 	 * W(ipf_global)->R(hook_family)->W(hei_lock) (this code path) vs
138 	 * R(hook_family)->R(hei_lock)->R(ipf_global) (active hook running)
139 	 */
140 	RWLOCK_EXIT(&ifs->ifs_ipf_global);
141 
142 #define	UNDO_HOOK(_f, _b, _e, _h)					\
143 	do {								\
144 		if (ifs->_f != NULL) {					\
145 			if (ifs->_b) {					\
146 				ifs->_b = (net_hook_unregister(ifs->_f,	\
147 					   _e, ifs->_h) != 0);		\
148 				if (!ifs->_b) {				\
149 					hook_free(ifs->_h);		\
150 					ifs->_h = NULL;			\
151 				}					\
152 			} else if (ifs->_h != NULL) {			\
153 				hook_free(ifs->_h);			\
154 				ifs->_h = NULL;				\
155 			}						\
156 		}							\
157 		_NOTE(CONSTCOND)					\
158 	} while (0)
159 
160 	/*
161 	 * Remove IPv6 Hooks
162 	 */
163 	if (ifs->ifs_ipf_ipv6 != NULL) {
164 		UNDO_HOOK(ifs_ipf_ipv6, ifs_hook6_physical_in,
165 			  NH_PHYSICAL_IN, ifs_ipfhook6_in);
166 		UNDO_HOOK(ifs_ipf_ipv6, ifs_hook6_physical_out,
167 			  NH_PHYSICAL_OUT, ifs_ipfhook6_out);
168 		UNDO_HOOK(ifs_ipf_ipv6, ifs_hook6_nic_events,
169 			  NH_NIC_EVENTS, ifs_ipfhook6_nicevents);
170 		UNDO_HOOK(ifs_ipf_ipv6, ifs_hook6_loopback_in,
171 			  NH_LOOPBACK_IN, ifs_ipfhook6_loop_in);
172 		UNDO_HOOK(ifs_ipf_ipv6, ifs_hook6_loopback_out,
173 			  NH_LOOPBACK_OUT, ifs_ipfhook6_loop_out);
174 
175 		if (net_protocol_release(ifs->ifs_ipf_ipv6) != 0)
176 			goto detach_failed;
177 		ifs->ifs_ipf_ipv6 = NULL;
178         }
179 
180 	/*
181 	 * Remove IPv4 Hooks
182 	 */
183 	if (ifs->ifs_ipf_ipv4 != NULL) {
184 		UNDO_HOOK(ifs_ipf_ipv4, ifs_hook4_physical_in,
185 			  NH_PHYSICAL_IN, ifs_ipfhook4_in);
186 		UNDO_HOOK(ifs_ipf_ipv4, ifs_hook4_physical_out,
187 			  NH_PHYSICAL_OUT, ifs_ipfhook4_out);
188 		UNDO_HOOK(ifs_ipf_ipv4, ifs_hook4_nic_events,
189 			  NH_NIC_EVENTS, ifs_ipfhook4_nicevents);
190 		UNDO_HOOK(ifs_ipf_ipv4, ifs_hook4_loopback_in,
191 			  NH_LOOPBACK_IN, ifs_ipfhook4_loop_in);
192 		UNDO_HOOK(ifs_ipf_ipv4, ifs_hook4_loopback_out,
193 			  NH_LOOPBACK_OUT, ifs_ipfhook4_loop_out);
194 
195 		if (net_protocol_release(ifs->ifs_ipf_ipv4) != 0)
196 			goto detach_failed;
197 		ifs->ifs_ipf_ipv4 = NULL;
198 	}
199 
200 #undef UNDO_HOOK
201 
202 #ifdef	IPFDEBUG
203 	cmn_err(CE_CONT, "ipldetach()\n");
204 #endif
205 
206 	WRITE_ENTER(&ifs->ifs_ipf_global);
207 	fr_deinitialise(ifs);
208 
209 	(void) frflush(IPL_LOGIPF, 0, FR_INQUE|FR_OUTQUE|FR_INACTIVE, ifs);
210 	(void) frflush(IPL_LOGIPF, 0, FR_INQUE|FR_OUTQUE, ifs);
211 
212 	if (ifs->ifs_ipf_locks_done == 1) {
213 		MUTEX_DESTROY(&ifs->ifs_ipf_timeoutlock);
214 		MUTEX_DESTROY(&ifs->ifs_ipf_rw);
215 		RW_DESTROY(&ifs->ifs_ipf_tokens);
216 		RW_DESTROY(&ifs->ifs_ipf_ipidfrag);
217 		ifs->ifs_ipf_locks_done = 0;
218 	}
219 
220 	if (ifs->ifs_hook4_physical_in || ifs->ifs_hook4_physical_out ||
221 	    ifs->ifs_hook4_nic_events || ifs->ifs_hook4_loopback_in ||
222 	    ifs->ifs_hook4_loopback_out || ifs->ifs_hook6_nic_events ||
223 	    ifs->ifs_hook6_physical_in || ifs->ifs_hook6_physical_out ||
224 	    ifs->ifs_hook6_loopback_in || ifs->ifs_hook6_loopback_out)
225 		return -1;
226 
227 	return 0;
228 
229 detach_failed:
230 	WRITE_ENTER(&ifs->ifs_ipf_global);
231 	return -1;
232 }
233 
234 int iplattach(ifs)
235 ipf_stack_t *ifs;
236 {
237 #if SOLARIS2 < 10
238 	int i;
239 #endif
240 	netid_t id = ifs->ifs_netid;
241 
242 #ifdef	IPFDEBUG
243 	cmn_err(CE_CONT, "iplattach()\n");
244 #endif
245 
246 	ASSERT(rw_read_locked(&ifs->ifs_ipf_global.ipf_lk) == 0);
247 	ifs->ifs_fr_flags = IPF_LOGGING;
248 #ifdef _KERNEL
249 	ifs->ifs_fr_update_ipid = 0;
250 #else
251 	ifs->ifs_fr_update_ipid = 1;
252 #endif
253 	ifs->ifs_fr_minttl = 4;
254 	ifs->ifs_fr_icmpminfragmtu = 68;
255 #if defined(IPFILTER_DEFAULT_BLOCK)
256 	ifs->ifs_fr_pass = FR_BLOCK|FR_NOMATCH;
257 #else
258 	ifs->ifs_fr_pass = (IPF_DEFAULT_PASS)|FR_NOMATCH;
259 #endif
260 
261 	bzero((char *)ifs->ifs_frcache, sizeof(ifs->ifs_frcache));
262 	MUTEX_INIT(&ifs->ifs_ipf_rw, "ipf rw mutex");
263 	MUTEX_INIT(&ifs->ifs_ipf_timeoutlock, "ipf timeout lock mutex");
264 	RWLOCK_INIT(&ifs->ifs_ipf_ipidfrag, "ipf IP NAT-Frag rwlock");
265 	RWLOCK_INIT(&ifs->ifs_ipf_tokens, "ipf token rwlock");
266 	ifs->ifs_ipf_locks_done = 1;
267 
268 	if (fr_initialise(ifs) < 0)
269 		return -1;
270 
271 	HOOK_INIT(ifs->ifs_ipfhook4_nicevents, ipf_nic_event_v4,
272 		  "ipfilter_hook4_nicevents", ifs);
273 	HOOK_INIT(ifs->ifs_ipfhook4_in, ipf_hook4_in,
274 		  "ipfilter_hook4_in", ifs);
275 	HOOK_INIT(ifs->ifs_ipfhook4_out, ipf_hook4_out,
276 		  "ipfilter_hook4_out", ifs);
277 	HOOK_INIT(ifs->ifs_ipfhook4_loop_in, ipf_hook4_loop_in,
278 		  "ipfilter_hook4_loop_in", ifs);
279 	HOOK_INIT(ifs->ifs_ipfhook4_loop_out, ipf_hook4_loop_out,
280 		  "ipfilter_hook4_loop_out", ifs);
281 
282 	/*
283 	 * If we hold this lock over all of the net_hook_register calls, we
284 	 * can cause a deadlock to occur with the following lock ordering:
285 	 * W(ipf_global)->R(hook_family)->W(hei_lock) (this code path) vs
286 	 * R(hook_family)->R(hei_lock)->R(ipf_global) (packet path)
287 	 */
288 	RWLOCK_EXIT(&ifs->ifs_ipf_global);
289 
290 	/*
291 	 * Add IPv4 hooks
292 	 */
293 	ifs->ifs_ipf_ipv4 = net_protocol_lookup(id, NHF_INET);
294 	if (ifs->ifs_ipf_ipv4 == NULL)
295 		goto hookup_failed;
296 
297 	ifs->ifs_hook4_nic_events = (net_hook_register(ifs->ifs_ipf_ipv4,
298 	    NH_NIC_EVENTS, ifs->ifs_ipfhook4_nicevents) == 0);
299 	if (!ifs->ifs_hook4_nic_events)
300 		goto hookup_failed;
301 
302 	ifs->ifs_hook4_physical_in = (net_hook_register(ifs->ifs_ipf_ipv4,
303 	    NH_PHYSICAL_IN, ifs->ifs_ipfhook4_in) == 0);
304 	if (!ifs->ifs_hook4_physical_in)
305 		goto hookup_failed;
306 
307 	ifs->ifs_hook4_physical_out = (net_hook_register(ifs->ifs_ipf_ipv4,
308 	    NH_PHYSICAL_OUT, ifs->ifs_ipfhook4_out) == 0);
309 	if (!ifs->ifs_hook4_physical_out)
310 		goto hookup_failed;
311 
312 	if (ifs->ifs_ipf_loopback) {
313 		ifs->ifs_hook4_loopback_in = (net_hook_register(
314 		    ifs->ifs_ipf_ipv4, NH_LOOPBACK_IN,
315 		    ifs->ifs_ipfhook4_loop_in) == 0);
316 		if (!ifs->ifs_hook4_loopback_in)
317 			goto hookup_failed;
318 
319 		ifs->ifs_hook4_loopback_out = (net_hook_register(
320 		    ifs->ifs_ipf_ipv4, NH_LOOPBACK_OUT,
321 		    ifs->ifs_ipfhook4_loop_out) == 0);
322 		if (!ifs->ifs_hook4_loopback_out)
323 			goto hookup_failed;
324 	}
325 	/*
326 	 * Add IPv6 hooks
327 	 */
328 	ifs->ifs_ipf_ipv6 = net_protocol_lookup(id, NHF_INET6);
329 	if (ifs->ifs_ipf_ipv6 == NULL)
330 		goto hookup_failed;
331 
332 	HOOK_INIT(ifs->ifs_ipfhook6_nicevents, ipf_nic_event_v6,
333 		  "ipfilter_hook6_nicevents", ifs);
334 	HOOK_INIT(ifs->ifs_ipfhook6_in, ipf_hook6_in,
335 		  "ipfilter_hook6_in", ifs);
336 	HOOK_INIT(ifs->ifs_ipfhook6_out, ipf_hook6_out,
337 		  "ipfilter_hook6_out", ifs);
338 	HOOK_INIT(ifs->ifs_ipfhook6_loop_in, ipf_hook6_loop_in,
339 		  "ipfilter_hook6_loop_in", ifs);
340 	HOOK_INIT(ifs->ifs_ipfhook6_loop_out, ipf_hook6_loop_out,
341 		  "ipfilter_hook6_loop_out", ifs);
342 
343 	ifs->ifs_hook6_nic_events = (net_hook_register(ifs->ifs_ipf_ipv6,
344 	    NH_NIC_EVENTS, ifs->ifs_ipfhook6_nicevents) == 0);
345 	if (!ifs->ifs_hook6_nic_events)
346 		goto hookup_failed;
347 
348 	ifs->ifs_hook6_physical_in = (net_hook_register(ifs->ifs_ipf_ipv6,
349 	    NH_PHYSICAL_IN, ifs->ifs_ipfhook6_in) == 0);
350 	if (!ifs->ifs_hook6_physical_in)
351 		goto hookup_failed;
352 
353 	ifs->ifs_hook6_physical_out = (net_hook_register(ifs->ifs_ipf_ipv6,
354 	    NH_PHYSICAL_OUT, ifs->ifs_ipfhook6_out) == 0);
355 	if (!ifs->ifs_hook6_physical_out)
356 		goto hookup_failed;
357 
358 	if (ifs->ifs_ipf_loopback) {
359 		ifs->ifs_hook6_loopback_in = (net_hook_register(
360 		    ifs->ifs_ipf_ipv6, NH_LOOPBACK_IN,
361 		    ifs->ifs_ipfhook6_loop_in) == 0);
362 		if (!ifs->ifs_hook6_loopback_in)
363 			goto hookup_failed;
364 
365 		ifs->ifs_hook6_loopback_out = (net_hook_register(
366 		    ifs->ifs_ipf_ipv6, NH_LOOPBACK_OUT,
367 		    ifs->ifs_ipfhook6_loop_out) == 0);
368 		if (!ifs->ifs_hook6_loopback_out)
369 			goto hookup_failed;
370 	}
371 
372 	/*
373 	 * Reacquire ipf_global, now it is safe.
374 	 */
375 	WRITE_ENTER(&ifs->ifs_ipf_global);
376 
377 /* Do not use private interface ip_params_arr[] in Solaris 10 */
378 #if SOLARIS2 < 10
379 
380 #if SOLARIS2 >= 8
381 	ip_forwarding = &ip_g_forward;
382 #endif
383 	/*
384 	 * XXX - There is no terminator for this array, so it is not possible
385 	 * to tell if what we are looking for is missing and go off the end
386 	 * of the array.
387 	 */
388 
389 #if SOLARIS2 <= 8
390 	for (i = 0; ; i++) {
391 		if (!strcmp(ip_param_arr[i].ip_param_name, "ip_def_ttl")) {
392 			ip_ttl_ptr = &ip_param_arr[i].ip_param_value;
393 		} else if (!strcmp(ip_param_arr[i].ip_param_name,
394 			    "ip_path_mtu_discovery")) {
395 			ip_mtudisc = &ip_param_arr[i].ip_param_value;
396 		}
397 #if SOLARIS2 < 8
398 		else if (!strcmp(ip_param_arr[i].ip_param_name,
399 			    "ip_forwarding")) {
400 			ip_forwarding = &ip_param_arr[i].ip_param_value;
401 		}
402 #else
403 		else if (!strcmp(ip_param_arr[i].ip_param_name,
404 			    "ip6_forwarding")) {
405 			ip6_forwarding = &ip_param_arr[i].ip_param_value;
406 		}
407 #endif
408 
409 		if (ip_mtudisc != NULL && ip_ttl_ptr != NULL &&
410 #if SOLARIS2 >= 8
411 		    ip6_forwarding != NULL &&
412 #endif
413 		    ip_forwarding != NULL)
414 			break;
415 	}
416 #endif
417 
418 	if (ifs->ifs_fr_control_forwarding & 1) {
419 		if (ip_forwarding != NULL)
420 			*ip_forwarding = 1;
421 #if SOLARIS2 >= 8
422 		if (ip6_forwarding != NULL)
423 			*ip6_forwarding = 1;
424 #endif
425 	}
426 
427 #endif
428 
429 	return 0;
430 hookup_failed:
431 	WRITE_ENTER(&ifs->ifs_ipf_global);
432 	return -1;
433 }
434 
435 static	int	fr_setipfloopback(set, ifs)
436 int set;
437 ipf_stack_t *ifs;
438 {
439 	if (ifs->ifs_ipf_ipv4 == NULL || ifs->ifs_ipf_ipv6 == NULL)
440 		return EFAULT;
441 
442 	if (set && !ifs->ifs_ipf_loopback) {
443 		ifs->ifs_ipf_loopback = 1;
444 
445 		ifs->ifs_hook4_loopback_in = (net_hook_register(
446 		    ifs->ifs_ipf_ipv4, NH_LOOPBACK_IN,
447 		    ifs->ifs_ipfhook4_loop_in) == 0);
448 		if (!ifs->ifs_hook4_loopback_in)
449 			return EINVAL;
450 
451 		ifs->ifs_hook4_loopback_out = (net_hook_register(
452 		    ifs->ifs_ipf_ipv4, NH_LOOPBACK_OUT,
453 		    ifs->ifs_ipfhook4_loop_out) == 0);
454 		if (!ifs->ifs_hook4_loopback_out)
455 			return EINVAL;
456 
457 		ifs->ifs_hook6_loopback_in = (net_hook_register(
458 		    ifs->ifs_ipf_ipv6, NH_LOOPBACK_IN,
459 		    ifs->ifs_ipfhook6_loop_in) == 0);
460 		if (!ifs->ifs_hook6_loopback_in)
461 			return EINVAL;
462 
463 		ifs->ifs_hook6_loopback_out = (net_hook_register(
464 		    ifs->ifs_ipf_ipv6, NH_LOOPBACK_OUT,
465 		    ifs->ifs_ipfhook6_loop_out) == 0);
466 		if (!ifs->ifs_hook6_loopback_out)
467 			return EINVAL;
468 
469 	} else if (!set && ifs->ifs_ipf_loopback) {
470 		ifs->ifs_ipf_loopback = 0;
471 
472 		ifs->ifs_hook4_loopback_in =
473 		    (net_hook_unregister(ifs->ifs_ipf_ipv4,
474 		    NH_LOOPBACK_IN, ifs->ifs_ipfhook4_loop_in) != 0);
475 		if (ifs->ifs_hook4_loopback_in)
476 			return EBUSY;
477 
478 		ifs->ifs_hook4_loopback_out =
479 		    (net_hook_unregister(ifs->ifs_ipf_ipv4,
480 		    NH_LOOPBACK_OUT, ifs->ifs_ipfhook4_loop_out) != 0);
481 		if (ifs->ifs_hook4_loopback_out)
482 			return EBUSY;
483 
484 		ifs->ifs_hook6_loopback_in =
485 		    (net_hook_unregister(ifs->ifs_ipf_ipv6,
486 		    NH_LOOPBACK_IN, ifs->ifs_ipfhook4_loop_in) != 0);
487 		if (ifs->ifs_hook6_loopback_in)
488 			return EBUSY;
489 
490 		ifs->ifs_hook6_loopback_out =
491 		    (net_hook_unregister(ifs->ifs_ipf_ipv6,
492 		    NH_LOOPBACK_OUT, ifs->ifs_ipfhook6_loop_out) != 0);
493 		if (ifs->ifs_hook6_loopback_out)
494 			return EBUSY;
495 	}
496 	return 0;
497 }
498 
499 
500 /*
501  * Filter ioctl interface.
502  */
503 /*ARGSUSED*/
504 int iplioctl(dev, cmd, data, mode, cp, rp)
505 dev_t dev;
506 int cmd;
507 #if SOLARIS2 >= 7
508 intptr_t data;
509 #else
510 int *data;
511 #endif
512 int mode;
513 cred_t *cp;
514 int *rp;
515 {
516 	int error = 0, tmp;
517 	friostat_t fio;
518 	minor_t unit;
519 	u_int enable;
520 	ipf_stack_t *ifs;
521 
522 #ifdef	IPFDEBUG
523 	cmn_err(CE_CONT, "iplioctl(%x,%x,%x,%d,%x,%d)\n",
524 		dev, cmd, data, mode, cp, rp);
525 #endif
526 	unit = getminor(dev);
527 	if (IPL_LOGMAX < unit)
528 		return ENXIO;
529 
530         /*
531 	 * As we're calling ipf_find_stack in user space, from a given zone
532 	 * to find the stack pointer for this zone, there is no need to have
533 	 * a hold/refence count here.
534 	 */
535 	ifs = ipf_find_stack(crgetzoneid(cp));
536 	ASSERT(ifs != NULL);
537 
538 	if (ifs->ifs_fr_running <= 0) {
539 		if (unit != IPL_LOGIPF) {
540 			return EIO;
541 		}
542 		if (cmd != SIOCIPFGETNEXT && cmd != SIOCIPFGET &&
543 		    cmd != SIOCIPFSET && cmd != SIOCFRENB &&
544 		    cmd != SIOCGETFS && cmd != SIOCGETFF) {
545 			return EIO;
546 		}
547 	}
548 
549 	READ_ENTER(&ifs->ifs_ipf_global);
550 	if (ifs->ifs_fr_enable_active != 0) {
551 		RWLOCK_EXIT(&ifs->ifs_ipf_global);
552 		return EBUSY;
553 	}
554 
555 	error = fr_ioctlswitch(unit, (caddr_t)data, cmd, mode, crgetuid(cp),
556 			       curproc, ifs);
557 	if (error != -1) {
558 		RWLOCK_EXIT(&ifs->ifs_ipf_global);
559 		return error;
560 	}
561 	error = 0;
562 
563 	switch (cmd)
564 	{
565 	case SIOCFRENB :
566 		if (!(mode & FWRITE))
567 			error = EPERM;
568 		else {
569 			error = COPYIN((caddr_t)data, (caddr_t)&enable,
570 				       sizeof(enable));
571 			if (error != 0) {
572 				error = EFAULT;
573 				break;
574 			}
575 
576 			RWLOCK_EXIT(&ifs->ifs_ipf_global);
577 			WRITE_ENTER(&ifs->ifs_ipf_global);
578 
579 			/*
580 			 * We must recheck fr_enable_active here, since we've
581 			 * dropped ifs_ipf_global from R in order to get it
582 			 * exclusively.
583 			 */
584 			if (ifs->ifs_fr_enable_active == 0) {
585 				ifs->ifs_fr_enable_active = 1;
586 				error = fr_enableipf(ifs, enable);
587 				ifs->ifs_fr_enable_active = 0;
588 			}
589 		}
590 		break;
591 	case SIOCIPFSET :
592 		if (!(mode & FWRITE)) {
593 			error = EPERM;
594 			break;
595 		}
596 		/* FALLTHRU */
597 	case SIOCIPFGETNEXT :
598 	case SIOCIPFGET :
599 		error = fr_ipftune(cmd, (void *)data, ifs);
600 		break;
601 	case SIOCSETFF :
602 		if (!(mode & FWRITE))
603 			error = EPERM;
604 		else {
605 			error = COPYIN((caddr_t)data,
606 				       (caddr_t)&ifs->ifs_fr_flags,
607 				       sizeof(ifs->ifs_fr_flags));
608 			if (error != 0)
609 				error = EFAULT;
610 		}
611 		break;
612 	case SIOCIPFLP :
613 		error = COPYIN((caddr_t)data, (caddr_t)&tmp,
614 			       sizeof(tmp));
615 		if (error != 0)
616 			error = EFAULT;
617 		else
618 			error = fr_setipfloopback(tmp, ifs);
619 		break;
620 	case SIOCGETFF :
621 		error = COPYOUT((caddr_t)&ifs->ifs_fr_flags, (caddr_t)data,
622 				sizeof(ifs->ifs_fr_flags));
623 		if (error != 0)
624 			error = EFAULT;
625 		break;
626 	case SIOCFUNCL :
627 		error = fr_resolvefunc((void *)data);
628 		break;
629 	case SIOCINAFR :
630 	case SIOCRMAFR :
631 	case SIOCADAFR :
632 	case SIOCZRLST :
633 		if (!(mode & FWRITE))
634 			error = EPERM;
635 		else
636 			error = frrequest(unit, cmd, (caddr_t)data,
637 					  ifs->ifs_fr_active, 1, ifs);
638 		break;
639 	case SIOCINIFR :
640 	case SIOCRMIFR :
641 	case SIOCADIFR :
642 		if (!(mode & FWRITE))
643 			error = EPERM;
644 		else
645 			error = frrequest(unit, cmd, (caddr_t)data,
646 					  1 - ifs->ifs_fr_active, 1, ifs);
647 		break;
648 	case SIOCSWAPA :
649 		if (!(mode & FWRITE))
650 			error = EPERM;
651 		else {
652 			WRITE_ENTER(&ifs->ifs_ipf_mutex);
653 			bzero((char *)ifs->ifs_frcache,
654 			    sizeof (ifs->ifs_frcache));
655 			error = COPYOUT((caddr_t)&ifs->ifs_fr_active,
656 					(caddr_t)data,
657 					sizeof(ifs->ifs_fr_active));
658 			if (error != 0)
659 				error = EFAULT;
660 			else
661 				ifs->ifs_fr_active = 1 - ifs->ifs_fr_active;
662 			RWLOCK_EXIT(&ifs->ifs_ipf_mutex);
663 		}
664 		break;
665 	case SIOCGETFS :
666 		fr_getstat(&fio, ifs);
667 		error = fr_outobj((void *)data, &fio, IPFOBJ_IPFSTAT);
668 		break;
669 	case SIOCFRZST :
670 		if (!(mode & FWRITE))
671 			error = EPERM;
672 		else
673 			error = fr_zerostats((caddr_t)data, ifs);
674 		break;
675 	case	SIOCIPFFL :
676 		if (!(mode & FWRITE))
677 			error = EPERM;
678 		else {
679 			error = COPYIN((caddr_t)data, (caddr_t)&tmp,
680 				       sizeof(tmp));
681 			if (!error) {
682 				tmp = frflush(unit, 4, tmp, ifs);
683 				error = COPYOUT((caddr_t)&tmp, (caddr_t)data,
684 						sizeof(tmp));
685 				if (error != 0)
686 					error = EFAULT;
687 			} else
688 				error = EFAULT;
689 		}
690 		break;
691 #ifdef USE_INET6
692 	case	SIOCIPFL6 :
693 		if (!(mode & FWRITE))
694 			error = EPERM;
695 		else {
696 			error = COPYIN((caddr_t)data, (caddr_t)&tmp,
697 				       sizeof(tmp));
698 			if (!error) {
699 				tmp = frflush(unit, 6, tmp, ifs);
700 				error = COPYOUT((caddr_t)&tmp, (caddr_t)data,
701 						sizeof(tmp));
702 				if (error != 0)
703 					error = EFAULT;
704 			} else
705 				error = EFAULT;
706 		}
707 		break;
708 #endif
709 	case SIOCSTLCK :
710 		error = COPYIN((caddr_t)data, (caddr_t)&tmp, sizeof(tmp));
711 		if (error == 0) {
712 			ifs->ifs_fr_state_lock = tmp;
713 			ifs->ifs_fr_nat_lock = tmp;
714 			ifs->ifs_fr_frag_lock = tmp;
715 			ifs->ifs_fr_auth_lock = tmp;
716 		} else
717 			error = EFAULT;
718 	break;
719 #ifdef	IPFILTER_LOG
720 	case	SIOCIPFFB :
721 		if (!(mode & FWRITE))
722 			error = EPERM;
723 		else {
724 			tmp = ipflog_clear(unit, ifs);
725 			error = COPYOUT((caddr_t)&tmp, (caddr_t)data,
726 				       sizeof(tmp));
727 			if (error)
728 				error = EFAULT;
729 		}
730 		break;
731 #endif /* IPFILTER_LOG */
732 	case SIOCFRSYN :
733 		if (!(mode & FWRITE))
734 			error = EPERM;
735 		else {
736 			RWLOCK_EXIT(&ifs->ifs_ipf_global);
737 			WRITE_ENTER(&ifs->ifs_ipf_global);
738 
739 			frsync(IPFSYNC_RESYNC, 0, NULL, NULL, ifs);
740 			fr_natifpsync(IPFSYNC_RESYNC, 0, NULL, NULL, ifs);
741 			fr_nataddrsync(0, NULL, NULL, ifs);
742 			fr_statesync(IPFSYNC_RESYNC, 0, NULL, NULL, ifs);
743 			error = 0;
744 		}
745 		break;
746 	case SIOCGFRST :
747 		error = fr_outobj((void *)data, fr_fragstats(ifs),
748 				  IPFOBJ_FRAGSTAT);
749 		break;
750 	case FIONREAD :
751 #ifdef	IPFILTER_LOG
752 		tmp = (int)ifs->ifs_iplused[IPL_LOGIPF];
753 
754 		error = COPYOUT((caddr_t)&tmp, (caddr_t)data, sizeof(tmp));
755 		if (error != 0)
756 			error = EFAULT;
757 #endif
758 		break;
759 	case SIOCIPFITER :
760 		error = ipf_frruleiter((caddr_t)data, crgetuid(cp),
761 				       curproc, ifs);
762 		break;
763 
764 	case SIOCGENITER :
765 		error = ipf_genericiter((caddr_t)data, crgetuid(cp),
766 					curproc, ifs);
767 		break;
768 
769 	case SIOCIPFDELTOK :
770 		error = BCOPYIN((caddr_t)data, (caddr_t)&tmp, sizeof(tmp));
771 		if (error != 0) {
772 			error = EFAULT;
773 		} else {
774 			error = ipf_deltoken(tmp, crgetuid(cp), curproc, ifs);
775 		}
776 		break;
777 
778 	default :
779 #ifdef	IPFDEBUG
780 		cmn_err(CE_NOTE, "Unknown: cmd 0x%x data %p",
781 			cmd, (void *)data);
782 #endif
783 		error = EINVAL;
784 		break;
785 	}
786 	RWLOCK_EXIT(&ifs->ifs_ipf_global);
787 	return error;
788 }
789 
790 
791 static int fr_enableipf(ifs, enable)
792 ipf_stack_t *ifs;
793 int enable;
794 {
795 	int error;
796 
797 	if (!enable) {
798 		error = ipldetach(ifs);
799 		if (error == 0)
800 			ifs->ifs_fr_running = -1;
801 		return error;
802 	}
803 
804 	if (ifs->ifs_fr_running > 0)
805 		return 0;
806 
807 	error = iplattach(ifs);
808 	if (error == 0) {
809 		if (ifs->ifs_fr_timer_id == NULL) {
810 			int hz = drv_usectohz(500000);
811 
812 			ifs->ifs_fr_timer_id = timeout(fr_slowtimer,
813 						       (void *)ifs,
814 						       hz);
815 		}
816 		ifs->ifs_fr_running = 1;
817 	} else {
818 		(void) ipldetach(ifs);
819 	}
820 	return error;
821 }
822 
823 
824 phy_if_t get_unit(name, v, ifs)
825 char *name;
826 int v;
827 ipf_stack_t *ifs;
828 {
829 	net_handle_t nif;
830 
831   	if (v == 4)
832  		nif = ifs->ifs_ipf_ipv4;
833   	else if (v == 6)
834  		nif = ifs->ifs_ipf_ipv6;
835   	else
836  		return 0;
837 
838  	return (net_phylookup(nif, name));
839 }
840 
841 /*
842  * routines below for saving IP headers to buffer
843  */
844 /*ARGSUSED*/
845 int iplopen(devp, flags, otype, cred)
846 dev_t *devp;
847 int flags, otype;
848 cred_t *cred;
849 {
850 	minor_t min = getminor(*devp);
851 
852 #ifdef	IPFDEBUG
853 	cmn_err(CE_CONT, "iplopen(%x,%x,%x,%x)\n", devp, flags, otype, cred);
854 #endif
855 	if (!(otype & OTYP_CHR))
856 		return ENXIO;
857 
858 	min = (IPL_LOGMAX < min) ? ENXIO : 0;
859 	return min;
860 }
861 
862 
863 /*ARGSUSED*/
864 int iplclose(dev, flags, otype, cred)
865 dev_t dev;
866 int flags, otype;
867 cred_t *cred;
868 {
869 	minor_t	min = getminor(dev);
870 
871 #ifdef	IPFDEBUG
872 	cmn_err(CE_CONT, "iplclose(%x,%x,%x,%x)\n", dev, flags, otype, cred);
873 #endif
874 
875 	min = (IPL_LOGMAX < min) ? ENXIO : 0;
876 	return min;
877 }
878 
879 #ifdef	IPFILTER_LOG
880 /*
881  * iplread/ipllog
882  * both of these must operate with at least splnet() lest they be
883  * called during packet processing and cause an inconsistancy to appear in
884  * the filter lists.
885  */
886 /*ARGSUSED*/
887 int iplread(dev, uio, cp)
888 dev_t dev;
889 register struct uio *uio;
890 cred_t *cp;
891 {
892 	ipf_stack_t *ifs;
893 	int ret;
894 
895         /*
896 	 * As we're calling ipf_find_stack in user space, from a given zone
897 	 * to find the stack pointer for this zone, there is no need to have
898 	 * a hold/refence count here.
899 	 */
900 	ifs = ipf_find_stack(crgetzoneid(cp));
901 	ASSERT(ifs != NULL);
902 
903 # ifdef	IPFDEBUG
904 	cmn_err(CE_CONT, "iplread(%x,%x,%x)\n", dev, uio, cp);
905 # endif
906 
907 	if (ifs->ifs_fr_running < 1) {
908 		return EIO;
909 	}
910 
911 # ifdef	IPFILTER_SYNC
912 	if (getminor(dev) == IPL_LOGSYNC) {
913 		return ipfsync_read(uio);
914 	}
915 # endif
916 
917 	ret = ipflog_read(getminor(dev), uio, ifs);
918 	return ret;
919 }
920 #endif /* IPFILTER_LOG */
921 
922 
923 /*
924  * iplread/ipllog
925  * both of these must operate with at least splnet() lest they be
926  * called during packet processing and cause an inconsistancy to appear in
927  * the filter lists.
928  */
929 int iplwrite(dev, uio, cp)
930 dev_t dev;
931 register struct uio *uio;
932 cred_t *cp;
933 {
934 	ipf_stack_t *ifs;
935 
936         /*
937 	 * As we're calling ipf_find_stack in user space, from a given zone
938 	 * to find the stack pointer for this zone, there is no need to have
939 	 * a hold/refence count here.
940 	 */
941 	ifs = ipf_find_stack(crgetzoneid(cp));
942 	ASSERT(ifs != NULL);
943 
944 #ifdef	IPFDEBUG
945 	cmn_err(CE_CONT, "iplwrite(%x,%x,%x)\n", dev, uio, cp);
946 #endif
947 
948 	if (ifs->ifs_fr_running < 1) {
949 		return EIO;
950 	}
951 
952 #ifdef	IPFILTER_SYNC
953 	if (getminor(dev) == IPL_LOGSYNC)
954 		return ipfsync_write(uio);
955 #endif /* IPFILTER_SYNC */
956 	dev = dev;	/* LINT */
957 	uio = uio;	/* LINT */
958 	cp = cp;	/* LINT */
959 	return ENXIO;
960 }
961 
962 
963 /*
964  * fr_send_reset - this could conceivably be a call to tcp_respond(), but that
965  * requires a large amount of setting up and isn't any more efficient.
966  */
967 int fr_send_reset(fin)
968 fr_info_t *fin;
969 {
970 	tcphdr_t *tcp, *tcp2;
971 	int tlen, hlen;
972 	mblk_t *m;
973 #ifdef	USE_INET6
974 	ip6_t *ip6;
975 #endif
976 	ip_t *ip;
977 
978 	tcp = fin->fin_dp;
979 	if (tcp->th_flags & TH_RST)
980 		return -1;
981 
982 #ifndef	IPFILTER_CKSUM
983 	if (fr_checkl4sum(fin) == -1)
984 		return -1;
985 #endif
986 
987 	tlen = (tcp->th_flags & (TH_SYN|TH_FIN)) ? 1 : 0;
988 #ifdef	USE_INET6
989 	if (fin->fin_v == 6)
990 		hlen = sizeof(ip6_t);
991 	else
992 #endif
993 		hlen = sizeof(ip_t);
994 	hlen += sizeof(*tcp2);
995 	if ((m = (mblk_t *)allocb(hlen + 64, BPRI_HI)) == NULL)
996 		return -1;
997 
998 	m->b_rptr += 64;
999 	MTYPE(m) = M_DATA;
1000 	m->b_wptr = m->b_rptr + hlen;
1001 	ip = (ip_t *)m->b_rptr;
1002 	bzero((char *)ip, hlen);
1003 	tcp2 = (struct tcphdr *)(m->b_rptr + hlen - sizeof(*tcp2));
1004 	tcp2->th_dport = tcp->th_sport;
1005 	tcp2->th_sport = tcp->th_dport;
1006 	if (tcp->th_flags & TH_ACK) {
1007 		tcp2->th_seq = tcp->th_ack;
1008 		tcp2->th_flags = TH_RST;
1009 	} else {
1010 		tcp2->th_ack = ntohl(tcp->th_seq);
1011 		tcp2->th_ack += tlen;
1012 		tcp2->th_ack = htonl(tcp2->th_ack);
1013 		tcp2->th_flags = TH_RST|TH_ACK;
1014 	}
1015 	tcp2->th_off = sizeof(struct tcphdr) >> 2;
1016 
1017 	ip->ip_v = fin->fin_v;
1018 #ifdef	USE_INET6
1019 	if (fin->fin_v == 6) {
1020 		ip6 = (ip6_t *)m->b_rptr;
1021 		ip6->ip6_flow = ((ip6_t *)fin->fin_ip)->ip6_flow;
1022 		ip6->ip6_src = fin->fin_dst6.in6;
1023 		ip6->ip6_dst = fin->fin_src6.in6;
1024 		ip6->ip6_plen = htons(sizeof(*tcp));
1025 		ip6->ip6_nxt = IPPROTO_TCP;
1026 		tcp2->th_sum = fr_cksum(m, (ip_t *)ip6, IPPROTO_TCP, tcp2);
1027 	} else
1028 #endif
1029 	{
1030 		ip->ip_src.s_addr = fin->fin_daddr;
1031 		ip->ip_dst.s_addr = fin->fin_saddr;
1032 		ip->ip_id = fr_nextipid(fin);
1033 		ip->ip_hl = sizeof(*ip) >> 2;
1034 		ip->ip_p = IPPROTO_TCP;
1035 		ip->ip_len = sizeof(*ip) + sizeof(*tcp);
1036 		ip->ip_tos = fin->fin_ip->ip_tos;
1037 		tcp2->th_sum = fr_cksum(m, ip, IPPROTO_TCP, tcp2);
1038 	}
1039 	return fr_send_ip(fin, m, &m);
1040 }
1041 
1042 /*
1043  * Function:	fr_send_ip
1044  * Returns:	 0: success
1045  *		-1: failed
1046  * Parameters:
1047  *	fin: packet information
1048  *	m: the message block where ip head starts
1049  *
1050  * Send a new packet through the IP stack.
1051  *
1052  * For IPv4 packets, ip_len must be in host byte order, and ip_v,
1053  * ip_ttl, ip_off, and ip_sum are ignored (filled in by this
1054  * function).
1055  *
1056  * For IPv6 packets, ip6_flow, ip6_vfc, and ip6_hlim are filled
1057  * in by this function.
1058  *
1059  * All other portions of the packet must be in on-the-wire format.
1060  */
1061 /*ARGSUSED*/
1062 static int fr_send_ip(fin, m, mpp)
1063 fr_info_t *fin;
1064 mblk_t *m, **mpp;
1065 {
1066 	qpktinfo_t qpi, *qpip;
1067 	fr_info_t fnew;
1068 	ip_t *ip;
1069 	int i, hlen;
1070 	ipf_stack_t *ifs = fin->fin_ifs;
1071 
1072 	ip = (ip_t *)m->b_rptr;
1073 	bzero((char *)&fnew, sizeof(fnew));
1074 
1075 #ifdef	USE_INET6
1076 	if (fin->fin_v == 6) {
1077 		ip6_t *ip6;
1078 
1079 		ip6 = (ip6_t *)ip;
1080 		ip6->ip6_vfc = 0x60;
1081 		ip6->ip6_hlim = 127;
1082 		fnew.fin_v = 6;
1083 		hlen = sizeof(*ip6);
1084 		fnew.fin_plen = ntohs(ip6->ip6_plen) + hlen;
1085 	} else
1086 #endif
1087 	{
1088 		fnew.fin_v = 4;
1089 #if SOLARIS2 >= 10
1090 		ip->ip_ttl = 255;
1091 		if (net_getpmtuenabled(ifs->ifs_ipf_ipv4) == 1)
1092 			ip->ip_off = htons(IP_DF);
1093 #else
1094 		if (ip_ttl_ptr != NULL)
1095 			ip->ip_ttl = (u_char)(*ip_ttl_ptr);
1096 		else
1097 			ip->ip_ttl = 63;
1098 		if (ip_mtudisc != NULL)
1099 			ip->ip_off = htons(*ip_mtudisc ? IP_DF : 0);
1100 		else
1101 			ip->ip_off = htons(IP_DF);
1102 #endif
1103 		/*
1104 		 * The dance with byte order and ip_len/ip_off is because in
1105 		 * fr_fastroute, it expects them to be in host byte order but
1106 		 * ipf_cksum expects them to be in network byte order.
1107 		 */
1108 		ip->ip_len = htons(ip->ip_len);
1109 		ip->ip_sum = ipf_cksum((u_short *)ip, sizeof(*ip));
1110 		ip->ip_len = ntohs(ip->ip_len);
1111 		ip->ip_off = ntohs(ip->ip_off);
1112 		hlen = sizeof(*ip);
1113 		fnew.fin_plen = ip->ip_len;
1114 	}
1115 
1116 	qpip = fin->fin_qpi;
1117 	qpi.qpi_off = 0;
1118 	qpi.qpi_ill = qpip->qpi_ill;
1119 	qpi.qpi_m = m;
1120 	qpi.qpi_data = ip;
1121 	fnew.fin_qpi = &qpi;
1122 	fnew.fin_ifp = fin->fin_ifp;
1123 	fnew.fin_flx = FI_NOCKSUM;
1124 	fnew.fin_m = m;
1125 	fnew.fin_qfm = m;
1126 	fnew.fin_ip = ip;
1127 	fnew.fin_mp = mpp;
1128 	fnew.fin_hlen = hlen;
1129 	fnew.fin_dp = (char *)ip + hlen;
1130 	fnew.fin_ifs = fin->fin_ifs;
1131 	(void) fr_makefrip(hlen, ip, &fnew);
1132 
1133 	i = fr_fastroute(m, mpp, &fnew, NULL);
1134 	return i;
1135 }
1136 
1137 
1138 int fr_send_icmp_err(type, fin, dst)
1139 int type;
1140 fr_info_t *fin;
1141 int dst;
1142 {
1143 	struct in_addr dst4;
1144 	struct icmp *icmp;
1145 	qpktinfo_t *qpi;
1146 	int hlen, code;
1147 	phy_if_t phy;
1148 	u_short sz;
1149 #ifdef	USE_INET6
1150 	mblk_t *mb;
1151 #endif
1152 	mblk_t *m;
1153 #ifdef	USE_INET6
1154 	ip6_t *ip6;
1155 #endif
1156 	ip_t *ip;
1157 	ipf_stack_t *ifs = fin->fin_ifs;
1158 
1159 	if ((type < 0) || (type > ICMP_MAXTYPE))
1160 		return -1;
1161 
1162 	code = fin->fin_icode;
1163 #ifdef USE_INET6
1164 	if ((code < 0) || (code >= ICMP_MAX_UNREACH))
1165 		return -1;
1166 #endif
1167 
1168 #ifndef	IPFILTER_CKSUM
1169 	if (fr_checkl4sum(fin) == -1)
1170 		return -1;
1171 #endif
1172 
1173 	qpi = fin->fin_qpi;
1174 
1175 #ifdef	USE_INET6
1176 	mb = fin->fin_qfm;
1177 
1178 	if (fin->fin_v == 6) {
1179 		sz = sizeof(ip6_t);
1180 		sz += MIN(mb->b_wptr - mb->b_rptr, 512);
1181 		hlen = sizeof(ip6_t);
1182 		type = icmptoicmp6types[type];
1183 		if (type == ICMP6_DST_UNREACH)
1184 			code = icmptoicmp6unreach[code];
1185 	} else
1186 #endif
1187 	{
1188 		if ((fin->fin_p == IPPROTO_ICMP) &&
1189 		    !(fin->fin_flx & FI_SHORT))
1190 			switch (ntohs(fin->fin_data[0]) >> 8)
1191 			{
1192 			case ICMP_ECHO :
1193 			case ICMP_TSTAMP :
1194 			case ICMP_IREQ :
1195 			case ICMP_MASKREQ :
1196 				break;
1197 			default :
1198 				return 0;
1199 			}
1200 
1201 		sz = sizeof(ip_t) * 2;
1202 		sz += 8;		/* 64 bits of data */
1203 		hlen = sizeof(ip_t);
1204 	}
1205 
1206 	sz += offsetof(struct icmp, icmp_ip);
1207 	if ((m = (mblk_t *)allocb((size_t)sz + 64, BPRI_HI)) == NULL)
1208 		return -1;
1209 	MTYPE(m) = M_DATA;
1210 	m->b_rptr += 64;
1211 	m->b_wptr = m->b_rptr + sz;
1212 	bzero((char *)m->b_rptr, (size_t)sz);
1213 	ip = (ip_t *)m->b_rptr;
1214 	ip->ip_v = fin->fin_v;
1215 	icmp = (struct icmp *)(m->b_rptr + hlen);
1216 	icmp->icmp_type = type & 0xff;
1217 	icmp->icmp_code = code & 0xff;
1218 	phy = (phy_if_t)qpi->qpi_ill;
1219 	if (type == ICMP_UNREACH && (phy != 0) &&
1220 	    fin->fin_icode == ICMP_UNREACH_NEEDFRAG)
1221 		icmp->icmp_nextmtu = net_getmtu(ifs->ifs_ipf_ipv4, phy,0 );
1222 
1223 #ifdef	USE_INET6
1224 	if (fin->fin_v == 6) {
1225 		struct in6_addr dst6;
1226 		int csz;
1227 
1228 		if (dst == 0) {
1229 			ipf_stack_t *ifs = fin->fin_ifs;
1230 
1231 			if (fr_ifpaddr(6, FRI_NORMAL, (void *)phy,
1232 				       (void *)&dst6, NULL, ifs) == -1) {
1233 				FREE_MB_T(m);
1234 				return -1;
1235 			}
1236 		} else
1237 			dst6 = fin->fin_dst6.in6;
1238 
1239 		csz = sz;
1240 		sz -= sizeof(ip6_t);
1241 		ip6 = (ip6_t *)m->b_rptr;
1242 		ip6->ip6_flow = ((ip6_t *)fin->fin_ip)->ip6_flow;
1243 		ip6->ip6_plen = htons((u_short)sz);
1244 		ip6->ip6_nxt = IPPROTO_ICMPV6;
1245 		ip6->ip6_src = dst6;
1246 		ip6->ip6_dst = fin->fin_src6.in6;
1247 		sz -= offsetof(struct icmp, icmp_ip);
1248 		bcopy((char *)mb->b_rptr, (char *)&icmp->icmp_ip, sz);
1249 		icmp->icmp_cksum = csz - sizeof(ip6_t);
1250 	} else
1251 #endif
1252 	{
1253 		ip->ip_hl = sizeof(*ip) >> 2;
1254 		ip->ip_p = IPPROTO_ICMP;
1255 		ip->ip_id = fin->fin_ip->ip_id;
1256 		ip->ip_tos = fin->fin_ip->ip_tos;
1257 		ip->ip_len = (u_short)sz;
1258 		if (dst == 0) {
1259 			ipf_stack_t *ifs = fin->fin_ifs;
1260 
1261 			if (fr_ifpaddr(4, FRI_NORMAL, (void *)phy,
1262 				       (void *)&dst4, NULL, ifs) == -1) {
1263 				FREE_MB_T(m);
1264 				return -1;
1265 			}
1266 		} else {
1267 			dst4 = fin->fin_dst;
1268 		}
1269 		ip->ip_src = dst4;
1270 		ip->ip_dst = fin->fin_src;
1271 		bcopy((char *)fin->fin_ip, (char *)&icmp->icmp_ip,
1272 		      sizeof(*fin->fin_ip));
1273 		bcopy((char *)fin->fin_ip + fin->fin_hlen,
1274 		      (char *)&icmp->icmp_ip + sizeof(*fin->fin_ip), 8);
1275 		icmp->icmp_ip.ip_len = htons(icmp->icmp_ip.ip_len);
1276 		icmp->icmp_ip.ip_off = htons(icmp->icmp_ip.ip_off);
1277 		icmp->icmp_cksum = ipf_cksum((u_short *)icmp,
1278 					     sz - sizeof(ip_t));
1279 	}
1280 
1281 	/*
1282 	 * Need to exit out of these so we don't recursively call rw_enter
1283 	 * from fr_qout.
1284 	 */
1285 	return fr_send_ip(fin, m, &m);
1286 }
1287 
1288 #include <sys/time.h>
1289 #include <sys/varargs.h>
1290 
1291 #ifndef _KERNEL
1292 #include <stdio.h>
1293 #endif
1294 
1295 /*
1296  * Return the first IP Address associated with an interface
1297  * For IPv6, we walk through the list of logical interfaces and return
1298  * the address of the first one that isn't a link-local interface.
1299  * We can't assume that it is :1 because another link-local address
1300  * may have been assigned there.
1301  */
1302 /*ARGSUSED*/
1303 int fr_ifpaddr(v, atype, ifptr, inp, inpmask, ifs)
1304 int v, atype;
1305 void *ifptr;
1306 struct in_addr  *inp, *inpmask;
1307 ipf_stack_t *ifs;
1308 {
1309 	struct sockaddr_in6 v6addr[2];
1310 	struct sockaddr_in v4addr[2];
1311 	net_ifaddr_t type[2];
1312 	net_handle_t net_data;
1313 	phy_if_t phyif;
1314 	void *array;
1315 
1316 	switch (v)
1317 	{
1318 	case 4:
1319 		net_data = ifs->ifs_ipf_ipv4;
1320 		array = v4addr;
1321 		break;
1322 	case 6:
1323 		net_data = ifs->ifs_ipf_ipv6;
1324 		array = v6addr;
1325 		break;
1326 	default:
1327 		net_data = NULL;
1328 		break;
1329 	}
1330 
1331 	if (net_data == NULL)
1332 		return -1;
1333 
1334 	phyif = (phy_if_t)ifptr;
1335 
1336 	switch (atype)
1337 	{
1338 	case FRI_PEERADDR :
1339 		type[0] = NA_PEER;
1340 		break;
1341 
1342 	case FRI_BROADCAST :
1343 		type[0] = NA_BROADCAST;
1344 		break;
1345 
1346 	default :
1347 		type[0] = NA_ADDRESS;
1348 		break;
1349 	}
1350 
1351 	type[1] = NA_NETMASK;
1352 
1353 	if (v == 6) {
1354 		lif_if_t idx = 0;
1355 
1356 		do {
1357 			idx = net_lifgetnext(net_data, phyif, idx);
1358 			if (net_getlifaddr(net_data, phyif, idx, 2, type,
1359 					   array) < 0)
1360 				return -1;
1361 			if (!IN6_IS_ADDR_LINKLOCAL(&v6addr[0].sin6_addr) &&
1362 			    !IN6_IS_ADDR_MULTICAST(&v6addr[0].sin6_addr))
1363 				break;
1364 		} while (idx != 0);
1365 
1366 		if (idx == 0)
1367 			return -1;
1368 
1369 		return fr_ifpfillv6addr(atype, &v6addr[0], &v6addr[1],
1370 					inp, inpmask);
1371 	}
1372 
1373 	if (net_getlifaddr(net_data, phyif, 0, 2, type, array) < 0)
1374 		return -1;
1375 
1376 	return fr_ifpfillv4addr(atype, &v4addr[0], &v4addr[1], inp, inpmask);
1377 }
1378 
1379 
1380 u_32_t fr_newisn(fin)
1381 fr_info_t *fin;
1382 {
1383 	static int iss_seq_off = 0;
1384 	u_char hash[16];
1385 	u_32_t newiss;
1386 	MD5_CTX ctx;
1387 	ipf_stack_t *ifs = fin->fin_ifs;
1388 
1389 	/*
1390 	 * Compute the base value of the ISS.  It is a hash
1391 	 * of (saddr, sport, daddr, dport, secret).
1392 	 */
1393 	MD5Init(&ctx);
1394 
1395 	MD5Update(&ctx, (u_char *) &fin->fin_fi.fi_src,
1396 		  sizeof(fin->fin_fi.fi_src));
1397 	MD5Update(&ctx, (u_char *) &fin->fin_fi.fi_dst,
1398 		  sizeof(fin->fin_fi.fi_dst));
1399 	MD5Update(&ctx, (u_char *) &fin->fin_dat, sizeof(fin->fin_dat));
1400 
1401 	MD5Update(&ctx, ifs->ifs_ipf_iss_secret, sizeof(ifs->ifs_ipf_iss_secret));
1402 
1403 	MD5Final(hash, &ctx);
1404 
1405 	bcopy(hash, &newiss, sizeof(newiss));
1406 
1407 	/*
1408 	 * Now increment our "timer", and add it in to
1409 	 * the computed value.
1410 	 *
1411 	 * XXX Use `addin'?
1412 	 * XXX TCP_ISSINCR too large to use?
1413 	 */
1414 	iss_seq_off += 0x00010000;
1415 	newiss += iss_seq_off;
1416 	return newiss;
1417 }
1418 
1419 
1420 /* ------------------------------------------------------------------------ */
1421 /* Function:    fr_nextipid                                                 */
1422 /* Returns:     int - 0 == success, -1 == error (packet should be droppped) */
1423 /* Parameters:  fin(I) - pointer to packet information                      */
1424 /*                                                                          */
1425 /* Returns the next IPv4 ID to use for this packet.                         */
1426 /* ------------------------------------------------------------------------ */
1427 u_short fr_nextipid(fin)
1428 fr_info_t *fin;
1429 {
1430 	static u_short ipid = 0;
1431 	u_short id;
1432 	ipf_stack_t *ifs = fin->fin_ifs;
1433 
1434 	MUTEX_ENTER(&ifs->ifs_ipf_rw);
1435 	if (fin->fin_pktnum != 0) {
1436 		id = fin->fin_pktnum & 0xffff;
1437 	} else {
1438 		id = ipid++;
1439 	}
1440 	MUTEX_EXIT(&ifs->ifs_ipf_rw);
1441 
1442 	return id;
1443 }
1444 
1445 
1446 #ifndef IPFILTER_CKSUM
1447 /* ARGSUSED */
1448 #endif
1449 INLINE void fr_checkv4sum(fin)
1450 fr_info_t *fin;
1451 {
1452 #ifdef IPFILTER_CKSUM
1453 	if (fr_checkl4sum(fin) == -1)
1454 		fin->fin_flx |= FI_BAD;
1455 #endif
1456 }
1457 
1458 
1459 #ifdef USE_INET6
1460 # ifndef IPFILTER_CKSUM
1461 /* ARGSUSED */
1462 # endif
1463 INLINE void fr_checkv6sum(fin)
1464 fr_info_t *fin;
1465 {
1466 # ifdef IPFILTER_CKSUM
1467 	if (fr_checkl4sum(fin) == -1)
1468 		fin->fin_flx |= FI_BAD;
1469 # endif
1470 }
1471 #endif /* USE_INET6 */
1472 
1473 
1474 #if (SOLARIS2 < 7)
1475 void fr_slowtimer()
1476 #else
1477 /*ARGSUSED*/
1478 void fr_slowtimer __P((void *arg))
1479 #endif
1480 {
1481 	ipf_stack_t *ifs = arg;
1482 
1483 	READ_ENTER(&ifs->ifs_ipf_global);
1484 	if (ifs->ifs_fr_running != 1) {
1485 		ifs->ifs_fr_timer_id = NULL;
1486 		RWLOCK_EXIT(&ifs->ifs_ipf_global);
1487 		return;
1488 	}
1489 	ipf_expiretokens(ifs);
1490 	fr_fragexpire(ifs);
1491 	fr_timeoutstate(ifs);
1492 	fr_natexpire(ifs);
1493 	fr_authexpire(ifs);
1494 	ifs->ifs_fr_ticks++;
1495 	if (ifs->ifs_fr_running == 1)
1496 		ifs->ifs_fr_timer_id = timeout(fr_slowtimer, arg,
1497 		    drv_usectohz(500000));
1498 	else
1499 		ifs->ifs_fr_timer_id = NULL;
1500 	RWLOCK_EXIT(&ifs->ifs_ipf_global);
1501 }
1502 
1503 
1504 /* ------------------------------------------------------------------------ */
1505 /* Function:    fr_pullup                                                   */
1506 /* Returns:     NULL == pullup failed, else pointer to protocol header      */
1507 /* Parameters:  m(I)   - pointer to buffer where data packet starts         */
1508 /*              fin(I) - pointer to packet information                      */
1509 /*              len(I) - number of bytes to pullup                          */
1510 /*                                                                          */
1511 /* Attempt to move at least len bytes (from the start of the buffer) into a */
1512 /* single buffer for ease of access.  Operating system native functions are */
1513 /* used to manage buffers - if necessary.  If the entire packet ends up in  */
1514 /* a single buffer, set the FI_COALESCE flag even though fr_coalesce() has  */
1515 /* not been called.  Both fin_ip and fin_dp are updated before exiting _IF_ */
1516 /* and ONLY if the pullup succeeds.                                         */
1517 /*                                                                          */
1518 /* We assume that 'min' is a pointer to a buffer that is part of the chain  */
1519 /* of buffers that starts at *fin->fin_mp.                                  */
1520 /* ------------------------------------------------------------------------ */
1521 void *fr_pullup(min, fin, len)
1522 mb_t *min;
1523 fr_info_t *fin;
1524 int len;
1525 {
1526 	qpktinfo_t *qpi = fin->fin_qpi;
1527 	int out = fin->fin_out, dpoff, ipoff;
1528 	mb_t *m = min, *m1, *m2;
1529 	char *ip;
1530 	uint32_t start, stuff, end, value, flags;
1531 	ipf_stack_t *ifs = fin->fin_ifs;
1532 
1533 	if (m == NULL)
1534 		return NULL;
1535 
1536 	ip = (char *)fin->fin_ip;
1537 	if ((fin->fin_flx & FI_COALESCE) != 0)
1538 		return ip;
1539 
1540 	ipoff = fin->fin_ipoff;
1541 	if (fin->fin_dp != NULL)
1542 		dpoff = (char *)fin->fin_dp - (char *)ip;
1543 	else
1544 		dpoff = 0;
1545 
1546 	if (M_LEN(m) < len + ipoff) {
1547 
1548 		/*
1549 		 * pfil_precheck ensures the IP header is on a 32bit
1550 		 * aligned address so simply fail if that isn't currently
1551 		 * the case (should never happen).
1552 		 */
1553 		int inc = 0;
1554 
1555 		if (ipoff > 0) {
1556 			if ((ipoff & 3) != 0) {
1557 				inc = 4 - (ipoff & 3);
1558 				if (m->b_rptr - inc >= m->b_datap->db_base)
1559 					m->b_rptr -= inc;
1560 				else
1561 					inc = 0;
1562 			}
1563 		}
1564 
1565 		/*
1566 		 * XXX This is here as a work around for a bug with DEBUG
1567 		 * XXX Solaris kernels.  The problem is b_prev is used by IP
1568 		 * XXX code as a way to stash the phyint_index for a packet,
1569 		 * XXX this doesn't get reset by IP but freeb does an ASSERT()
1570 		 * XXX for both of these to be NULL.  See 6442390.
1571 		 */
1572 		m1 = m;
1573 		m2 = m->b_prev;
1574 
1575 		do {
1576 			m1->b_next = NULL;
1577 			m1->b_prev = NULL;
1578 			m1 = m1->b_cont;
1579 		} while (m1);
1580 
1581 		/*
1582 		 * Need to preserve checksum information by copying them
1583 		 * to newmp which heads the pulluped message.
1584 		 */
1585 		hcksum_retrieve(m, NULL, NULL, &start, &stuff, &end,
1586 		    &value, &flags);
1587 
1588 		if (pullupmsg(m, len + ipoff + inc) == 0) {
1589 			ATOMIC_INCL(ifs->ifs_frstats[out].fr_pull[1]);
1590 			FREE_MB_T(*fin->fin_mp);
1591 			*fin->fin_mp = NULL;
1592 			fin->fin_m = NULL;
1593 			fin->fin_ip = NULL;
1594 			fin->fin_dp = NULL;
1595 			qpi->qpi_data = NULL;
1596 			return NULL;
1597 		}
1598 
1599 		(void) hcksum_assoc(m, NULL, NULL, start, stuff, end,
1600 		    value, flags, 0);
1601 
1602 		m->b_prev = m2;
1603 		m->b_rptr += inc;
1604 		fin->fin_m = m;
1605 		ip = MTOD(m, char *) + ipoff;
1606 		qpi->qpi_data = ip;
1607 	}
1608 
1609 	ATOMIC_INCL(ifs->ifs_frstats[out].fr_pull[0]);
1610 	fin->fin_ip = (ip_t *)ip;
1611 	if (fin->fin_dp != NULL)
1612 		fin->fin_dp = (char *)fin->fin_ip + dpoff;
1613 
1614 	if (len == fin->fin_plen)
1615 		fin->fin_flx |= FI_COALESCE;
1616 	return ip;
1617 }
1618 
1619 
1620 /*
1621  * Function:	fr_verifysrc
1622  * Returns:	int (really boolean)
1623  * Parameters:	fin - packet information
1624  *
1625  * Check whether the packet has a valid source address for the interface on
1626  * which the packet arrived, implementing the "fr_chksrc" feature.
1627  * Returns true iff the packet's source address is valid.
1628  */
1629 int fr_verifysrc(fin)
1630 fr_info_t *fin;
1631 {
1632 	net_handle_t net_data_p;
1633 	phy_if_t phy_ifdata_routeto;
1634 	struct sockaddr	sin;
1635 	ipf_stack_t *ifs = fin->fin_ifs;
1636 
1637 	if (fin->fin_v == 4) {
1638 		net_data_p = ifs->ifs_ipf_ipv4;
1639 	} else if (fin->fin_v == 6) {
1640 		net_data_p = ifs->ifs_ipf_ipv6;
1641 	} else {
1642 		return (0);
1643 	}
1644 
1645 	/* Get the index corresponding to the if name */
1646 	sin.sa_family = (fin->fin_v == 4) ? AF_INET : AF_INET6;
1647 	bcopy(&fin->fin_saddr, &sin.sa_data, sizeof (struct in_addr));
1648 	phy_ifdata_routeto = net_routeto(net_data_p, &sin, NULL);
1649 
1650 	return (((phy_if_t)fin->fin_ifp == phy_ifdata_routeto) ? 1 : 0);
1651 }
1652 
1653 
1654 /*
1655  * Function:	fr_fastroute
1656  * Returns:	 0: success;
1657  *		-1: failed
1658  * Parameters:
1659  *	mb: the message block where ip head starts
1660  *	mpp: the pointer to the pointer of the orignal
1661  *		packet message
1662  *	fin: packet information
1663  *	fdp: destination interface information
1664  *	if it is NULL, no interface information provided.
1665  *
1666  * This function is for fastroute/to/dup-to rules. It calls
1667  * pfil_make_lay2_packet to search route, make lay-2 header
1668  * ,and identify output queue for the IP packet.
1669  * The destination address depends on the following conditions:
1670  * 1: for fastroute rule, fdp is passed in as NULL, so the
1671  *	destination address is the IP Packet's destination address
1672  * 2: for to/dup-to rule, if an ip address is specified after
1673  *	the interface name, this address is the as destination
1674  *	address. Otherwise IP Packet's destination address is used
1675  */
1676 int fr_fastroute(mb, mpp, fin, fdp)
1677 mblk_t *mb, **mpp;
1678 fr_info_t *fin;
1679 frdest_t *fdp;
1680 {
1681         net_handle_t net_data_p;
1682 	net_inject_t *inj;
1683 	mblk_t *mp = NULL;
1684 	frentry_t *fr = fin->fin_fr;
1685 	qpktinfo_t *qpi;
1686 	ip_t *ip;
1687 
1688 	struct sockaddr_in *sin;
1689 	struct sockaddr_in6 *sin6;
1690 	struct sockaddr *sinp;
1691 	ipf_stack_t *ifs = fin->fin_ifs;
1692 #ifndef	sparc
1693 	u_short __iplen, __ipoff;
1694 #endif
1695 
1696 	if (fin->fin_v == 4) {
1697 		net_data_p = ifs->ifs_ipf_ipv4;
1698 	} else if (fin->fin_v == 6) {
1699 		net_data_p = ifs->ifs_ipf_ipv6;
1700 	} else {
1701 		return (-1);
1702 	}
1703 
1704 	inj = net_inject_alloc(NETINFO_VERSION);
1705 	if (inj == NULL)
1706 		return -1;
1707 
1708 	ip = fin->fin_ip;
1709 	qpi = fin->fin_qpi;
1710 
1711 	/*
1712 	 * If this is a duplicate mblk then we want ip to point at that
1713 	 * data, not the original, if and only if it is already pointing at
1714 	 * the current mblk data.
1715 	 *
1716 	 * Otherwise, if it's not a duplicate, and we're not already pointing
1717 	 * at the current mblk data, then we want to ensure that the data
1718 	 * points at ip.
1719 	 */
1720 
1721 	if ((ip == (ip_t *)qpi->qpi_m->b_rptr) && (qpi->qpi_m != mb)) {
1722 		ip = (ip_t *)mb->b_rptr;
1723 	} else if ((qpi->qpi_m == mb) && (ip != (ip_t *)qpi->qpi_m->b_rptr)) {
1724 		qpi->qpi_m->b_rptr = (uchar_t *)ip;
1725 		qpi->qpi_off = 0;
1726 	}
1727 
1728 	/*
1729 	 * If there is another M_PROTO, we don't want it
1730 	 */
1731 	if (*mpp != mb) {
1732 		mp = unlinkb(*mpp);
1733 		freeb(*mpp);
1734 		*mpp = mp;
1735 	}
1736 
1737 	sinp = (struct sockaddr *)&inj->ni_addr;
1738 	sin = (struct sockaddr_in *)sinp;
1739 	sin6 = (struct sockaddr_in6 *)sinp;
1740 	bzero((char *)&inj->ni_addr, sizeof (inj->ni_addr));
1741 	inj->ni_addr.ss_family = (fin->fin_v == 4) ? AF_INET : AF_INET6;
1742 	inj->ni_packet = mb;
1743 
1744 	/*
1745 	 * In case we're here due to "to <if>" being used with
1746 	 * "keep state", check that we're going in the correct
1747 	 * direction.
1748 	 */
1749 	if (fdp != NULL) {
1750 		if ((fr != NULL) && (fdp->fd_ifp != NULL) &&
1751 			(fin->fin_rev != 0) && (fdp == &fr->fr_tif))
1752 			goto bad_fastroute;
1753 		inj->ni_physical = (phy_if_t)fdp->fd_ifp;
1754 		if (fin->fin_v == 4) {
1755 			sin->sin_addr = fdp->fd_ip;
1756 		} else {
1757 			sin6->sin6_addr = fdp->fd_ip6.in6;
1758 		}
1759 	} else {
1760 		if (fin->fin_v == 4) {
1761 			sin->sin_addr = ip->ip_dst;
1762 		} else {
1763 			sin6->sin6_addr = ((ip6_t *)ip)->ip6_dst;
1764 		}
1765 		inj->ni_physical = net_routeto(net_data_p, sinp, NULL);
1766 	}
1767 
1768 	/*
1769 	 * Clear the hardware checksum flags from packets that we are doing
1770 	 * input processing on as leaving them set will cause the outgoing
1771 	 * NIC (if it supports hardware checksum) to calculate them anew,
1772 	 * using the old (correct) checksums as the pseudo value to start
1773 	 * from.
1774 	 */
1775 	if (fin->fin_out == 0) {
1776 		DB_CKSUMFLAGS(mb) = 0;
1777 	}
1778 
1779 	*mpp = mb;
1780 
1781 	if (fin->fin_out == 0) {
1782 		void *saveifp;
1783 		u_32_t pass;
1784 
1785 		saveifp = fin->fin_ifp;
1786 		fin->fin_ifp = (void *)inj->ni_physical;
1787 		fin->fin_flx &= ~FI_STATE;
1788 		fin->fin_out = 1;
1789 		(void) fr_acctpkt(fin, &pass);
1790 		fin->fin_fr = NULL;
1791 		if (!fr || !(fr->fr_flags & FR_RETMASK))
1792 			(void) fr_checkstate(fin, &pass);
1793 		if (fr_checknatout(fin, NULL) == -1)
1794 			goto bad_fastroute;
1795 		fin->fin_out = 0;
1796 		fin->fin_ifp = saveifp;
1797 	}
1798 #ifndef	sparc
1799 	if (fin->fin_v == 4) {
1800 		__iplen = (u_short)ip->ip_len,
1801 		__ipoff = (u_short)ip->ip_off;
1802 
1803 		ip->ip_len = htons(__iplen);
1804 		ip->ip_off = htons(__ipoff);
1805 	}
1806 #endif
1807 
1808 	if (net_data_p) {
1809 		if (net_inject(net_data_p, NI_DIRECT_OUT, inj) < 0) {
1810 			net_inject_free(inj);
1811 			return (-1);
1812 		}
1813 	}
1814 
1815 	ifs->ifs_fr_frouteok[0]++;
1816 	net_inject_free(inj);
1817 	return 0;
1818 bad_fastroute:
1819 	net_inject_free(inj);
1820 	freemsg(mb);
1821 	ifs->ifs_fr_frouteok[1]++;
1822 	return -1;
1823 }
1824 
1825 
1826 /* ------------------------------------------------------------------------ */
1827 /* Function:    ipf_hook4_out                                               */
1828 /* Returns:     int - 0 == packet ok, else problem, free packet if not done */
1829 /* Parameters:  event(I)     - pointer to event                             */
1830 /*              info(I)      - pointer to hook information for firewalling  */
1831 /*                                                                          */
1832 /* Calling ipf_hook.                                                        */
1833 /* ------------------------------------------------------------------------ */
1834 /*ARGSUSED*/
1835 int ipf_hook4_out(hook_event_token_t token, hook_data_t info, void *arg)
1836 {
1837 	return ipf_hook(info, 1, 0, arg);
1838 }
1839 /*ARGSUSED*/
1840 int ipf_hook6_out(hook_event_token_t token, hook_data_t info, void *arg)
1841 {
1842 	return ipf_hook6(info, 1, 0, arg);
1843 }
1844 
1845 /* ------------------------------------------------------------------------ */
1846 /* Function:    ipf_hook4_in                                                */
1847 /* Returns:     int - 0 == packet ok, else problem, free packet if not done */
1848 /* Parameters:  event(I)     - pointer to event                             */
1849 /*              info(I)      - pointer to hook information for firewalling  */
1850 /*                                                                          */
1851 /* Calling ipf_hook.                                                        */
1852 /* ------------------------------------------------------------------------ */
1853 /*ARGSUSED*/
1854 int ipf_hook4_in(hook_event_token_t token, hook_data_t info, void *arg)
1855 {
1856 	return ipf_hook(info, 0, 0, arg);
1857 }
1858 /*ARGSUSED*/
1859 int ipf_hook6_in(hook_event_token_t token, hook_data_t info, void *arg)
1860 {
1861 	return ipf_hook6(info, 0, 0, arg);
1862 }
1863 
1864 
1865 /* ------------------------------------------------------------------------ */
1866 /* Function:    ipf_hook4_loop_out                                          */
1867 /* Returns:     int - 0 == packet ok, else problem, free packet if not done */
1868 /* Parameters:  event(I)     - pointer to event                             */
1869 /*              info(I)      - pointer to hook information for firewalling  */
1870 /*                                                                          */
1871 /* Calling ipf_hook.                                                        */
1872 /* ------------------------------------------------------------------------ */
1873 /*ARGSUSED*/
1874 int ipf_hook4_loop_out(hook_event_token_t token, hook_data_t info, void *arg)
1875 {
1876 	return ipf_hook(info, 1, FI_NOCKSUM, arg);
1877 }
1878 /*ARGSUSED*/
1879 int ipf_hook6_loop_out(hook_event_token_t token, hook_data_t info, void *arg)
1880 {
1881 	return ipf_hook6(info, 1, FI_NOCKSUM, arg);
1882 }
1883 
1884 /* ------------------------------------------------------------------------ */
1885 /* Function:    ipf_hook4_loop_in                                           */
1886 /* Returns:     int - 0 == packet ok, else problem, free packet if not done */
1887 /* Parameters:  event(I)     - pointer to event                             */
1888 /*              info(I)      - pointer to hook information for firewalling  */
1889 /*                                                                          */
1890 /* Calling ipf_hook.                                                        */
1891 /* ------------------------------------------------------------------------ */
1892 /*ARGSUSED*/
1893 int ipf_hook4_loop_in(hook_event_token_t token, hook_data_t info, void *arg)
1894 {
1895 	return ipf_hook(info, 0, FI_NOCKSUM, arg);
1896 }
1897 /*ARGSUSED*/
1898 int ipf_hook6_loop_in(hook_event_token_t token, hook_data_t info, void *arg)
1899 {
1900 	return ipf_hook6(info, 0, FI_NOCKSUM, arg);
1901 }
1902 
1903 /* ------------------------------------------------------------------------ */
1904 /* Function:    ipf_hook                                                    */
1905 /* Returns:     int - 0 == packet ok, else problem, free packet if not done */
1906 /* Parameters:  info(I)      - pointer to hook information for firewalling  */
1907 /*              out(I)       - whether packet is going in or out            */
1908 /*              loopback(I)  - whether packet is a loopback packet or not   */
1909 /*                                                                          */
1910 /* Stepping stone function between the IP mainline and IPFilter.  Extracts  */
1911 /* parameters out of the info structure and forms them up to be useful for  */
1912 /* calling ipfilter.                                                        */
1913 /* ------------------------------------------------------------------------ */
1914 int ipf_hook(hook_data_t info, int out, int loopback, void *arg)
1915 {
1916 	hook_pkt_event_t *fw;
1917 	ipf_stack_t *ifs;
1918 	qpktinfo_t qpi;
1919 	int rval, hlen;
1920 	u_short swap;
1921 	phy_if_t phy;
1922 	ip_t *ip;
1923 
1924 	ifs = arg;
1925 	fw = (hook_pkt_event_t *)info;
1926 
1927 	ASSERT(fw != NULL);
1928 	phy = (out == 0) ? fw->hpe_ifp : fw->hpe_ofp;
1929 
1930 	ip = fw->hpe_hdr;
1931 	swap = ntohs(ip->ip_len);
1932 	ip->ip_len = swap;
1933 	swap = ntohs(ip->ip_off);
1934 	ip->ip_off = swap;
1935 	hlen = IPH_HDR_LENGTH(ip);
1936 
1937 	qpi.qpi_m = fw->hpe_mb;
1938 	qpi.qpi_data = fw->hpe_hdr;
1939 	qpi.qpi_off = (char *)qpi.qpi_data - (char *)fw->hpe_mb->b_rptr;
1940 	qpi.qpi_ill = (void *)phy;
1941 	qpi.qpi_flags = fw->hpe_flags & (HPE_MULTICAST|HPE_BROADCAST);
1942 	if (qpi.qpi_flags)
1943 		qpi.qpi_flags |= FI_MBCAST;
1944 	qpi.qpi_flags |= loopback;
1945 
1946 	rval = fr_check(fw->hpe_hdr, hlen, qpi.qpi_ill, out,
1947 	    &qpi, fw->hpe_mp, ifs);
1948 
1949 	/* For fastroute cases, fr_check returns 0 with mp set to NULL */
1950 	if (rval == 0 && *(fw->hpe_mp) == NULL)
1951 		rval = 1;
1952 
1953 	/* Notify IP the packet mblk_t and IP header pointers. */
1954 	fw->hpe_mb = qpi.qpi_m;
1955 	fw->hpe_hdr = qpi.qpi_data;
1956 	if (rval == 0) {
1957 		ip = qpi.qpi_data;
1958 		swap = ntohs(ip->ip_len);
1959 		ip->ip_len = swap;
1960 		swap = ntohs(ip->ip_off);
1961 		ip->ip_off = swap;
1962 	}
1963 	return rval;
1964 
1965 }
1966 int ipf_hook6(hook_data_t info, int out, int loopback, void *arg)
1967 {
1968 	hook_pkt_event_t *fw;
1969 	int rval, hlen;
1970 	qpktinfo_t qpi;
1971 	phy_if_t phy;
1972 
1973 	fw = (hook_pkt_event_t *)info;
1974 
1975 	ASSERT(fw != NULL);
1976 	phy = (out == 0) ? fw->hpe_ifp : fw->hpe_ofp;
1977 
1978 	hlen = sizeof (ip6_t);
1979 
1980 	qpi.qpi_m = fw->hpe_mb;
1981 	qpi.qpi_data = fw->hpe_hdr;
1982 	qpi.qpi_off = (char *)qpi.qpi_data - (char *)fw->hpe_mb->b_rptr;
1983 	qpi.qpi_ill = (void *)phy;
1984 	qpi.qpi_flags = fw->hpe_flags & (HPE_MULTICAST|HPE_BROADCAST);
1985 	if (qpi.qpi_flags)
1986 		qpi.qpi_flags |= FI_MBCAST;
1987 	qpi.qpi_flags |= loopback;
1988 
1989 	rval = fr_check(fw->hpe_hdr, hlen, qpi.qpi_ill, out,
1990 	    &qpi, fw->hpe_mp, arg);
1991 
1992 	/* For fastroute cases, fr_check returns 0 with mp set to NULL */
1993 	if (rval == 0 && *(fw->hpe_mp) == NULL)
1994 		rval = 1;
1995 
1996 	/* Notify IP the packet mblk_t and IP header pointers. */
1997 	fw->hpe_mb = qpi.qpi_m;
1998 	fw->hpe_hdr = qpi.qpi_data;
1999 	return rval;
2000 
2001 }
2002 
2003 
2004 /* ------------------------------------------------------------------------ */
2005 /* Function:    ipf_nic_event_v4                                            */
2006 /* Returns:     int - 0 == no problems encountered                          */
2007 /* Parameters:  event(I)     - pointer to event                             */
2008 /*              info(I)      - pointer to information about a NIC event     */
2009 /*                                                                          */
2010 /* Function to receive asynchronous NIC events from IP                      */
2011 /* ------------------------------------------------------------------------ */
2012 /*ARGSUSED*/
2013 int ipf_nic_event_v4(hook_event_token_t event, hook_data_t info, void *arg)
2014 {
2015 	struct sockaddr_in *sin;
2016 	hook_nic_event_t *hn;
2017 	ipf_stack_t *ifs = arg;
2018 	void *new_ifp = NULL;
2019 
2020 	if (ifs->ifs_fr_running <= 0)
2021 		return (0);
2022 
2023 	hn = (hook_nic_event_t *)info;
2024 
2025 	switch (hn->hne_event)
2026 	{
2027 	case NE_PLUMB :
2028 		frsync(IPFSYNC_NEWIFP, 4, (void *)hn->hne_nic, hn->hne_data,
2029 		       ifs);
2030 		fr_natifpsync(IPFSYNC_NEWIFP, 4, (void *)hn->hne_nic,
2031 			      hn->hne_data, ifs);
2032 		fr_statesync(IPFSYNC_NEWIFP, 4, (void *)hn->hne_nic,
2033 			     hn->hne_data, ifs);
2034 		break;
2035 
2036 	case NE_UNPLUMB :
2037 		frsync(IPFSYNC_OLDIFP, 4, (void *)hn->hne_nic, NULL, ifs);
2038 		fr_natifpsync(IPFSYNC_OLDIFP, 4, (void *)hn->hne_nic, NULL,
2039 			      ifs);
2040 		fr_statesync(IPFSYNC_OLDIFP, 4, (void *)hn->hne_nic, NULL, ifs);
2041 		break;
2042 
2043 	case NE_ADDRESS_CHANGE :
2044 		/*
2045 		 * We only respond to events for logical interface 0 because
2046 		 * IPFilter only uses the first address given to a network
2047 		 * interface.  We check for hne_lif==1 because the netinfo
2048 		 * code maps adds 1 to the lif number so that it can return
2049 		 * 0 to indicate "no more lifs" when walking them.
2050 		 */
2051 		if (hn->hne_lif == 1) {
2052 			frsync(IPFSYNC_RESYNC, 4, (void *)hn->hne_nic, NULL,
2053 			    ifs);
2054 			sin = hn->hne_data;
2055 			fr_nataddrsync(4, (void *)hn->hne_nic, &sin->sin_addr,
2056 			    ifs);
2057 		}
2058 		break;
2059 
2060 #if SOLARIS2 >= 10
2061 	case NE_IFINDEX_CHANGE :
2062 		WRITE_ENTER(&ifs->ifs_ipf_mutex);
2063 
2064 		if (hn->hne_data != NULL) {
2065 			/*
2066 			 * The netinfo passes interface index as int (hne_data should be
2067 			 * handled as a pointer to int), which is always 32bit. We need to
2068 			 * convert it to void pointer here, since interfaces are
2069 			 * represented as pointers to void in IPF. The pointers are 64 bits
2070 			 * long on 64bit platforms. Doing something like
2071 			 *	(void *)((int) x)
2072 			 * will throw warning:
2073 			 *   "cast to pointer from integer of different size"
2074 			 * during 64bit compilation.
2075 			 *
2076 			 * The line below uses (size_t) to typecast int to
2077 			 * size_t, which might be 64bit/32bit (depending
2078 			 * on architecture). Once we have proper 64bit/32bit
2079 			 * type (size_t), we can safely convert it to void pointer.
2080 			 */
2081 			new_ifp = (void *)(size_t)*((int *)hn->hne_data);
2082 			fr_ifindexsync((void *)hn->hne_nic, new_ifp, ifs);
2083 			fr_natifindexsync((void *)hn->hne_nic, new_ifp, ifs);
2084 			fr_stateifindexsync((void *)hn->hne_nic, new_ifp, ifs);
2085 		}
2086 		RWLOCK_EXIT(&ifs->ifs_ipf_mutex);
2087 		break;
2088 #endif
2089 
2090 	default :
2091 		break;
2092 	}
2093 
2094 	return 0;
2095 }
2096 
2097 
2098 /* ------------------------------------------------------------------------ */
2099 /* Function:    ipf_nic_event_v6                                            */
2100 /* Returns:     int - 0 == no problems encountered                          */
2101 /* Parameters:  event(I)     - pointer to event                             */
2102 /*              info(I)      - pointer to information about a NIC event     */
2103 /*                                                                          */
2104 /* Function to receive asynchronous NIC events from IP                      */
2105 /* ------------------------------------------------------------------------ */
2106 /*ARGSUSED*/
2107 int ipf_nic_event_v6(hook_event_token_t event, hook_data_t info, void *arg)
2108 {
2109 	struct sockaddr_in6 *sin6;
2110 	hook_nic_event_t *hn;
2111 	ipf_stack_t *ifs = arg;
2112 	void *new_ifp = NULL;
2113 
2114 	if (ifs->ifs_fr_running <= 0)
2115 		return (0);
2116 
2117 	hn = (hook_nic_event_t *)info;
2118 
2119 	switch (hn->hne_event)
2120 	{
2121 	case NE_PLUMB :
2122 		frsync(IPFSYNC_NEWIFP, 6, (void *)hn->hne_nic,
2123 		       hn->hne_data, ifs);
2124 		fr_natifpsync(IPFSYNC_NEWIFP, 6, (void *)hn->hne_nic,
2125 			      hn->hne_data, ifs);
2126 		fr_statesync(IPFSYNC_NEWIFP, 6, (void *)hn->hne_nic,
2127 			     hn->hne_data, ifs);
2128 		break;
2129 
2130 	case NE_UNPLUMB :
2131 		frsync(IPFSYNC_OLDIFP, 6, (void *)hn->hne_nic, NULL, ifs);
2132 		fr_natifpsync(IPFSYNC_OLDIFP, 6, (void *)hn->hne_nic, NULL,
2133 			      ifs);
2134 		fr_statesync(IPFSYNC_OLDIFP, 6, (void *)hn->hne_nic, NULL, ifs);
2135 		break;
2136 
2137 	case NE_ADDRESS_CHANGE :
2138 		if (hn->hne_lif == 1) {
2139 			sin6 = hn->hne_data;
2140 			fr_nataddrsync(6, (void *)hn->hne_nic, &sin6->sin6_addr,
2141 				       ifs);
2142 		}
2143 		break;
2144 
2145 #if SOLARIS2 >= 10
2146 	case NE_IFINDEX_CHANGE :
2147 		WRITE_ENTER(&ifs->ifs_ipf_mutex);
2148 		if (hn->hne_data != NULL) {
2149 			/*
2150 			 * The netinfo passes interface index as int (hne_data should be
2151 			 * handled as a pointer to int), which is always 32bit. We need to
2152 			 * convert it to void pointer here, since interfaces are
2153 			 * represented as pointers to void in IPF. The pointers are 64 bits
2154 			 * long on 64bit platforms. Doing something like
2155 			 *	(void *)((int) x)
2156 			 * will throw warning:
2157 			 *   "cast to pointer from integer of different size"
2158 			 * during 64bit compilation.
2159 			 *
2160 			 * The line below uses (size_t) to typecast int to
2161 			 * size_t, which might be 64bit/32bit (depending
2162 			 * on architecture). Once we have proper 64bit/32bit
2163 			 * type (size_t), we can safely convert it to void pointer.
2164 			 */
2165 			new_ifp = (void *)(size_t)*((int *)hn->hne_data);
2166 			fr_ifindexsync((void *)hn->hne_nic, new_ifp, ifs);
2167 			fr_natifindexsync((void *)hn->hne_nic, new_ifp, ifs);
2168 			fr_stateifindexsync((void *)hn->hne_nic, new_ifp, ifs);
2169 		}
2170 		RWLOCK_EXIT(&ifs->ifs_ipf_mutex);
2171 		break;
2172 #endif
2173 
2174 	default :
2175 		break;
2176 	}
2177 
2178 	return 0;
2179 }
2180 
2181 /*
2182  * Functions fr_make_rst(), fr_make_icmp_v4(), fr_make_icmp_v6()
2183  * are needed in Solaris kernel only. We don't need them in
2184  * ipftest to pretend the ICMP/RST packet was sent as a response.
2185  */
2186 #if defined(_KERNEL) && (SOLARIS2 >= 10)
2187 /* ------------------------------------------------------------------------ */
2188 /* Function:    fr_make_rst                                                 */
2189 /* Returns:     int - 0 on success, -1 on failure			    */
2190 /* Parameters:  fin(I) - pointer to packet information                      */
2191 /*                                                                          */
2192 /* We must alter the original mblks passed to IPF from IP stack via	    */
2193 /* FW_HOOKS. FW_HOOKS interface is powerfull, but it has some limitations.  */
2194 /* IPF can basicaly do only these things with mblk representing the packet: */
2195 /*	leave it as it is (pass the packet)				    */
2196 /*                                                                          */
2197 /*	discard it (block the packet)					    */
2198 /*                                                                          */
2199 /*	alter it (i.e. NAT)						    */
2200 /*                                                                          */
2201 /* As you can see IPF can not simply discard the mblk and supply a new one  */
2202 /* instead to IP stack via FW_HOOKS.					    */
2203 /*                                                                          */
2204 /* The return-rst action for packets coming via NIC is handled as follows:  */
2205 /*	mblk with packet is discarded					    */
2206 /*                                                                          */
2207 /*	new mblk with RST response is constructed and injected to network   */
2208 /*                                                                          */
2209 /* IPF can't inject packets to loopback interface, this is just another	    */
2210 /* limitation we have to deal with here. The only option to send RST	    */
2211 /* response to offending TCP packet coming via loopback is to alter it.	    */
2212 /*									    */
2213 /* The fr_make_rst() function alters TCP SYN/FIN packet intercepted on	    */
2214 /* loopback interface into TCP RST packet. fin->fin_mp is pointer to	    */
2215 /* mblk L3 (IP) and L4 (TCP/UDP) packet headers.			    */
2216 /* ------------------------------------------------------------------------ */
2217 int fr_make_rst(fin)
2218 fr_info_t *fin;
2219 {
2220 	uint16_t tmp_port;
2221 	int rv = -1;
2222 	uint32_t old_ack;
2223 	tcphdr_t *tcp = NULL;
2224 	struct in_addr tmp_src;
2225 #ifdef USE_INET6
2226 	struct in6_addr	tmp_src6;
2227 #endif
2228 
2229 	ASSERT(fin->fin_p == IPPROTO_TCP);
2230 
2231 	/*
2232 	 * We do not need to adjust chksum, since it is not being checked by
2233 	 * Solaris IP stack for loopback clients.
2234 	 */
2235 	if ((fin->fin_v == 4) && (fin->fin_p == IPPROTO_TCP) &&
2236 	    ((tcp = (tcphdr_t *) fin->fin_dp) != NULL)) {
2237 
2238 		if (tcp->th_flags & (TH_SYN | TH_FIN)) {
2239 			/* Swap IPv4 addresses. */
2240 			tmp_src = fin->fin_ip->ip_src;
2241 			fin->fin_ip->ip_src = fin->fin_ip->ip_dst;
2242 			fin->fin_ip->ip_dst = tmp_src;
2243 
2244 			rv = 0;
2245 		}
2246 		else
2247 			tcp = NULL;
2248 	}
2249 #ifdef USE_INET6
2250 	else if ((fin->fin_v == 6) && (fin->fin_p == IPPROTO_TCP) &&
2251 	    ((tcp = (tcphdr_t *) fin->fin_dp) != NULL)) {
2252 		/*
2253 		 * We are relying on fact the next header is TCP, which is true
2254 		 * for regular TCP packets coming in over loopback.
2255 		 */
2256 		if (tcp->th_flags & (TH_SYN | TH_FIN)) {
2257 			/* Swap IPv6 addresses. */
2258 			tmp_src6 = fin->fin_ip6->ip6_src;
2259 			fin->fin_ip6->ip6_src = fin->fin_ip6->ip6_dst;
2260 			fin->fin_ip6->ip6_dst = tmp_src6;
2261 
2262 			rv = 0;
2263 		}
2264 		else
2265 			tcp = NULL;
2266 	}
2267 #endif
2268 
2269 	if (tcp != NULL) {
2270 		/*
2271 		 * Adjust TCP header:
2272 		 *	swap ports,
2273 		 *	set flags,
2274 		 *	set correct ACK number
2275 		 */
2276 		tmp_port = tcp->th_sport;
2277 		tcp->th_sport = tcp->th_dport;
2278 		tcp->th_dport = tmp_port;
2279 		old_ack = tcp->th_ack;
2280 		tcp->th_ack = htonl(ntohl(tcp->th_seq) + 1);
2281 		tcp->th_seq = old_ack;
2282 		tcp->th_flags = TH_RST | TH_ACK;
2283 	}
2284 
2285 	return (rv);
2286 }
2287 
2288 /* ------------------------------------------------------------------------ */
2289 /* Function:    fr_make_icmp_v4                                             */
2290 /* Returns:     int - 0 on success, -1 on failure			    */
2291 /* Parameters:  fin(I) - pointer to packet information                      */
2292 /*                                                                          */
2293 /* Please read comment at fr_make_icmp() wrapper function to get an idea    */
2294 /* what is going to happen here and why. Once you read the comment there,   */
2295 /* continue here with next paragraph.					    */
2296 /*									    */
2297 /* To turn IPv4 packet into ICMPv4 response packet, these things must	    */
2298 /* happen here:								    */
2299 /*	(1) Original mblk is copied (duplicated).			    */
2300 /*                                                                          */
2301 /*	(2) ICMP header is created.					    */
2302 /*                                                                          */
2303 /*	(3) Link ICMP header with copy of original mblk, we have ICMPv4	    */
2304 /*	    data ready then.						    */
2305 /*                                                                          */
2306 /*      (4) Swap IP addresses in original mblk and adjust IP header data.   */
2307 /*                                                                          */
2308 /*	(5) The mblk containing original packet is trimmed to contain IP    */
2309 /*	    header only and ICMP chksum is computed.			    */
2310 /*                                                                          */
2311 /*	(6) The ICMP header we have from (3) is linked to original mblk,    */
2312 /*	    which now contains new IP header. If original packet was spread */
2313 /*	    over several mblks, only the first mblk is kept.		    */
2314 /* ------------------------------------------------------------------------ */
2315 static int fr_make_icmp_v4(fin)
2316 fr_info_t *fin;
2317 {
2318 	struct in_addr tmp_src;
2319 	tcphdr_t *tcp;
2320 	struct icmp *icmp;
2321 	mblk_t *mblk_icmp;
2322 	mblk_t *mblk_ip;
2323 	size_t icmp_pld_len;	/* octets to append to ICMP header */
2324 	size_t orig_iphdr_len;	/* length of IP header only */
2325 	uint32_t sum;
2326 	uint16_t *buf;
2327 	int len;
2328 
2329 
2330 	if (fin->fin_v != 4)
2331 		return (-1);
2332 
2333 	/*
2334 	 * If we are dealing with TCP, then packet must be SYN/FIN to be routed
2335 	 * by IP stack. If it is not SYN/FIN, then we must drop it silently.
2336 	 */
2337 	tcp = (tcphdr_t *) fin->fin_dp;
2338 
2339 	if ((fin->fin_p == IPPROTO_TCP) &&
2340 	    ((tcp == NULL) || ((tcp->th_flags & (TH_SYN | TH_FIN)) == 0)))
2341 		return (-1);
2342 
2343 	/*
2344 	 * Step (1)
2345 	 *
2346 	 * Make copy of original mblk.
2347 	 *
2348 	 * We want to copy as much data as necessary, not less, not more.  The
2349 	 * ICMPv4 payload length for unreachable messages is:
2350 	 *	original IP header + 8 bytes of L4 (if there are any).
2351 	 *
2352 	 * We determine if there are at least 8 bytes of L4 data following IP
2353 	 * header first.
2354 	 */
2355 	icmp_pld_len = (fin->fin_dlen > ICMPERR_ICMPHLEN) ?
2356 		ICMPERR_ICMPHLEN : fin->fin_dlen;
2357 	/*
2358 	 * Since we don't want to copy more data than necessary, we must trim
2359 	 * the original mblk here.  The right way (STREAMish) would be to use
2360 	 * adjmsg() to trim it.  However we would have to calculate the length
2361 	 * argument for adjmsg() from pointers we already have here.
2362 	 *
2363 	 * Since we have pointers and offsets, it's faster and easier for
2364 	 * us to just adjust pointers by hand instead of using adjmsg().
2365 	 */
2366 	fin->fin_m->b_wptr = (unsigned char *) fin->fin_dp;
2367 	fin->fin_m->b_wptr += icmp_pld_len;
2368 	icmp_pld_len = fin->fin_m->b_wptr - (unsigned char *) fin->fin_ip;
2369 
2370 	/*
2371 	 * Also we don't want to copy any L2 stuff, which might precede IP
2372 	 * header, so we have have to set b_rptr to point to the start of IP
2373 	 * header.
2374 	 */
2375 	fin->fin_m->b_rptr += fin->fin_ipoff;
2376 	if ((mblk_ip = copyb(fin->fin_m)) == NULL)
2377 		return (-1);
2378 	fin->fin_m->b_rptr -= fin->fin_ipoff;
2379 
2380 	/*
2381 	 * Step (2)
2382 	 *
2383 	 * Create an ICMP header, which will be appened to original mblk later.
2384 	 * ICMP header is just another mblk.
2385 	 */
2386 	mblk_icmp = (mblk_t *) allocb(ICMPERR_ICMPHLEN, BPRI_HI);
2387 	if (mblk_icmp == NULL) {
2388 		FREE_MB_T(mblk_ip);
2389 		return (-1);
2390 	}
2391 
2392 	MTYPE(mblk_icmp) = M_DATA;
2393 	icmp = (struct icmp *) mblk_icmp->b_wptr;
2394 	icmp->icmp_type = ICMP_UNREACH;
2395 	icmp->icmp_code = fin->fin_icode & 0xFF;
2396 	icmp->icmp_void = 0;
2397 	icmp->icmp_cksum = 0;
2398 	mblk_icmp->b_wptr += ICMPERR_ICMPHLEN;
2399 
2400 	/*
2401 	 * Step (3)
2402 	 *
2403 	 * Complete ICMP packet - link ICMP header with L4 data from original
2404 	 * IP packet.
2405 	 */
2406 	linkb(mblk_icmp, mblk_ip);
2407 
2408 	/*
2409 	 * Step (4)
2410 	 *
2411 	 * Swap IP addresses and change IP header fields accordingly in
2412 	 * original IP packet.
2413 	 *
2414 	 * There is a rule option return-icmp as a dest for physical
2415 	 * interfaces. This option becomes useless for loopback, since IPF box
2416 	 * uses same address as a loopback destination. We ignore the option
2417 	 * here, the ICMP packet will always look like as it would have been
2418 	 * sent from the original destination host.
2419 	 */
2420 	tmp_src = fin->fin_ip->ip_src;
2421 	fin->fin_ip->ip_src = fin->fin_ip->ip_dst;
2422 	fin->fin_ip->ip_dst = tmp_src;
2423 	fin->fin_ip->ip_p = IPPROTO_ICMP;
2424 	fin->fin_ip->ip_sum = 0;
2425 
2426 	/*
2427 	 * Step (5)
2428 	 *
2429 	 * We trim the orignal mblk to hold IP header only.
2430 	 */
2431 	fin->fin_m->b_wptr = fin->fin_dp;
2432 	orig_iphdr_len = fin->fin_m->b_wptr -
2433 			    (fin->fin_m->b_rptr + fin->fin_ipoff);
2434 	fin->fin_ip->ip_len = htons(icmp_pld_len + ICMPERR_ICMPHLEN +
2435 			    orig_iphdr_len);
2436 
2437 	/*
2438 	 * ICMP chksum calculation. The data we are calculating chksum for are
2439 	 * spread over two mblks, therefore we have to use two for loops.
2440 	 *
2441 	 * First for loop computes chksum part for ICMP header.
2442 	 */
2443 	buf = (uint16_t *) icmp;
2444 	len = ICMPERR_ICMPHLEN;
2445 	for (sum = 0; len > 1; len -= 2)
2446 		sum += *buf++;
2447 
2448 	/*
2449 	 * Here we add chksum part for ICMP payload.
2450 	 */
2451 	len = icmp_pld_len;
2452 	buf = (uint16_t *) mblk_ip->b_rptr;
2453 	for (; len > 1; len -= 2)
2454 		sum += *buf++;
2455 
2456 	/*
2457 	 * Chksum is done.
2458 	 */
2459 	sum = (sum >> 16) + (sum & 0xffff);
2460 	sum += (sum >> 16);
2461 	icmp->icmp_cksum = ~sum;
2462 
2463 	/*
2464 	 * Step (6)
2465 	 *
2466 	 * Release all packet mblks, except the first one.
2467 	 */
2468 	if (fin->fin_m->b_cont != NULL) {
2469 		FREE_MB_T(fin->fin_m->b_cont);
2470 	}
2471 
2472 	/*
2473 	 * Append ICMP payload to first mblk, which already contains new IP
2474 	 * header.
2475 	 */
2476 	linkb(fin->fin_m, mblk_icmp);
2477 
2478 	return (0);
2479 }
2480 
2481 #ifdef USE_INET6
2482 /* ------------------------------------------------------------------------ */
2483 /* Function:    fr_make_icmp_v6                                             */
2484 /* Returns:     int - 0 on success, -1 on failure			    */
2485 /* Parameters:  fin(I) - pointer to packet information                      */
2486 /*									    */
2487 /* Please read comment at fr_make_icmp() wrapper function to get an idea    */
2488 /* what and why is going to happen here. Once you read the comment there,   */
2489 /* continue here with next paragraph.					    */
2490 /*									    */
2491 /* This function turns IPv6 packet (UDP, TCP, ...) into ICMPv6 response.    */
2492 /* The algorithm is fairly simple:					    */
2493 /*	1) We need to get copy of complete mblk.			    */
2494 /*									    */
2495 /*	2) New ICMPv6 header is created.				    */
2496 /*									    */
2497 /*	3) The copy of original mblk with packet is linked to ICMPv6	    */
2498 /*	   header.							    */
2499 /*									    */
2500 /*	4) The checksum must be adjusted.				    */
2501 /*									    */
2502 /*	5) IP addresses in original mblk are swapped and IP header data	    */
2503 /*	   are adjusted (protocol number).				    */
2504 /*									    */
2505 /*	6) Original mblk is trimmed to hold IPv6 header only, then it is    */
2506 /*	   linked with the ICMPv6 data we got from (3).			    */
2507 /* ------------------------------------------------------------------------ */
2508 static int fr_make_icmp_v6(fin)
2509 fr_info_t *fin;
2510 {
2511 	struct icmp6_hdr *icmp6;
2512 	tcphdr_t *tcp;
2513 	struct in6_addr	tmp_src6;
2514 	size_t icmp_pld_len;
2515 	mblk_t *mblk_ip, *mblk_icmp;
2516 
2517 	if (fin->fin_v != 6)
2518 		return (-1);
2519 
2520 	/*
2521 	 * If we are dealing with TCP, then packet must SYN/FIN to be routed by
2522 	 * IP stack. If it is not SYN/FIN, then we must drop it silently.
2523 	 */
2524 	tcp = (tcphdr_t *) fin->fin_dp;
2525 
2526 	if ((fin->fin_p == IPPROTO_TCP) &&
2527 	    ((tcp == NULL) || ((tcp->th_flags & (TH_SYN | TH_FIN)) == 0)))
2528 		return (-1);
2529 
2530 	/*
2531 	 * Step (1)
2532 	 *
2533 	 * We need to copy complete packet in case of IPv6, no trimming is
2534 	 * needed (except the L2 headers).
2535 	 */
2536 	icmp_pld_len = M_LEN(fin->fin_m);
2537 	fin->fin_m->b_rptr += fin->fin_ipoff;
2538 	if ((mblk_ip = copyb(fin->fin_m)) == NULL)
2539 		return (-1);
2540 	fin->fin_m->b_rptr -= fin->fin_ipoff;
2541 
2542 	/*
2543 	 * Step (2)
2544 	 *
2545 	 * Allocate and create ICMP header.
2546 	 */
2547 	mblk_icmp = (mblk_t *) allocb(sizeof (struct icmp6_hdr),
2548 			BPRI_HI);
2549 
2550 	if (mblk_icmp == NULL)
2551 		return (-1);
2552 
2553 	MTYPE(mblk_icmp) = M_DATA;
2554 	icmp6 =  (struct icmp6_hdr *) mblk_icmp->b_wptr;
2555 	icmp6->icmp6_type = ICMP6_DST_UNREACH;
2556 	icmp6->icmp6_code = fin->fin_icode & 0xFF;
2557 	icmp6->icmp6_data32[0] = 0;
2558 	mblk_icmp->b_wptr += sizeof (struct icmp6_hdr);
2559 
2560 	/*
2561 	 * Step (3)
2562 	 *
2563 	 * Link the copy of IP packet to ICMP header.
2564 	 */
2565 	linkb(mblk_icmp, mblk_ip);
2566 
2567 	/*
2568 	 * Step (4)
2569 	 *
2570 	 * Calculate chksum - this is much more easier task than in case of
2571 	 * IPv4  - ICMPv6 chksum only covers IP addresses, and payload length.
2572 	 * We are making compensation just for change of packet length.
2573 	 */
2574 	icmp6->icmp6_cksum = icmp_pld_len + sizeof (struct icmp6_hdr);
2575 
2576 	/*
2577 	 * Step (5)
2578 	 *
2579 	 * Swap IP addresses.
2580 	 */
2581 	tmp_src6 = fin->fin_ip6->ip6_src;
2582 	fin->fin_ip6->ip6_src = fin->fin_ip6->ip6_dst;
2583 	fin->fin_ip6->ip6_dst = tmp_src6;
2584 
2585 	/*
2586 	 * and adjust IP header data.
2587 	 */
2588 	fin->fin_ip6->ip6_nxt = IPPROTO_ICMPV6;
2589 	fin->fin_ip6->ip6_plen = htons(icmp_pld_len + sizeof (struct icmp6_hdr));
2590 
2591 	/*
2592 	 * Step (6)
2593 	 *
2594 	 * We must release all linked mblks from original packet and keep only
2595 	 * the first mblk with IP header to link ICMP data.
2596 	 */
2597 	fin->fin_m->b_wptr = (unsigned char *) fin->fin_ip6 + sizeof (ip6_t);
2598 
2599 	if (fin->fin_m->b_cont != NULL) {
2600 		FREE_MB_T(fin->fin_m->b_cont);
2601 	}
2602 
2603 	/*
2604 	 * Append ICMP payload to IP header.
2605 	 */
2606 	linkb(fin->fin_m, mblk_icmp);
2607 
2608 	return (0);
2609 }
2610 #endif	/* USE_INET6 */
2611 
2612 /* ------------------------------------------------------------------------ */
2613 /* Function:    fr_make_icmp                                                */
2614 /* Returns:     int - 0 on success, -1 on failure			    */
2615 /* Parameters:  fin(I) - pointer to packet information                      */
2616 /*                                                                          */
2617 /* We must alter the original mblks passed to IPF from IP stack via	    */
2618 /* FW_HOOKS. The reasons why we must alter packet are discussed within	    */
2619 /* comment at fr_make_rst() function.					    */
2620 /*									    */
2621 /* The fr_make_icmp() function acts as a wrapper, which passes the code	    */
2622 /* execution to	fr_make_icmp_v4() or fr_make_icmp_v6() depending on	    */
2623 /* protocol version. However there are some details, which are common to    */
2624 /* both IP versions. The details are going to be explained here.	    */
2625 /*                                                                          */
2626 /* The packet looks as follows:						    */
2627 /*    xxx | IP hdr | IP payload    ...	| 				    */
2628 /*    ^   ^        ^            	^				    */
2629 /*    |   |        |            	|				    */
2630 /*    |   |        |		fin_m->b_wptr = fin->fin_dp + fin->fin_dlen */
2631 /*    |   |        |							    */
2632 /*    |   |        `- fin_m->fin_dp (in case of IPv4 points to L4 header)   */
2633 /*    |   |								    */
2634 /*    |   `- fin_m->b_rptr + fin_ipoff (fin_ipoff is most likely 0 in case  */
2635 /*    |      of loopback)						    */
2636 /*    |   								    */
2637 /*    `- fin_m->b_rptr -  points to L2 header in case of physical NIC	    */
2638 /*                                                                          */
2639 /* All relevant IP headers are pulled up into the first mblk. It happened   */
2640 /* well in advance before the matching rule was found (the rule, which took */
2641 /* us here, to fr_make_icmp() function).				    */
2642 /*                                                                          */
2643 /* Both functions will turn packet passed in fin->fin_m mblk into a new	    */
2644 /* packet. New packet will be represented as chain of mblks.		    */
2645 /* orig mblk |- b_cont ---.						    */
2646 /*    ^                    `-> ICMP hdr |- b_cont--.			    */
2647 /*    |	                          ^	            `-> duped orig mblk	    */
2648 /*    |                           |				^	    */
2649 /*    `- The original mblk        |				|	    */
2650 /*       will be trimmed to       |				|	    */
2651 /*       to contain IP header     |				|	    */
2652 /*       only                     |				|	    */
2653 /*                                |				|	    */
2654 /*                                `- This is newly		|           */
2655 /*                                   allocated mblk to		|	    */
2656 /*                                   hold ICMPv6 data.		|	    */
2657 /*								|	    */
2658 /*								|	    */
2659 /*								|	    */
2660 /*	    This is the copy of original mblk, it will contain -'	    */
2661 /*	    orignal IP  packet in case of ICMPv6. In case of		    */
2662 /*	    ICMPv4 it will contain up to 8 bytes of IP payload		    */
2663 /*	    (TCP/UDP/L4) data from original packet.			    */
2664 /* ------------------------------------------------------------------------ */
2665 int fr_make_icmp(fin)
2666 fr_info_t *fin;
2667 {
2668 	int rv;
2669 
2670 	if (fin->fin_v == 4)
2671 		rv = fr_make_icmp_v4(fin);
2672 #ifdef USE_INET6
2673 	else if (fin->fin_v == 6)
2674 		rv = fr_make_icmp_v6(fin);
2675 #endif
2676 	else
2677 		rv = -1;
2678 
2679 	return (rv);
2680 }
2681 
2682 /* ------------------------------------------------------------------------ */
2683 /* Function:    fr_buf_sum						    */
2684 /* Returns:     unsigned int - sum of buffer buf			    */
2685 /* Parameters:  buf - pointer to buf we want to sum up			    */
2686 /*              len - length of buffer buf				    */
2687 /*                                                                          */
2688 /* Sums buffer buf. The result is used for chksum calculation. The buf	    */
2689 /* argument must be aligned.						    */
2690 /* ------------------------------------------------------------------------ */
2691 static uint32_t fr_buf_sum(buf, len)
2692 const void *buf;
2693 unsigned int len;
2694 {
2695 	uint32_t	sum = 0;
2696 	uint16_t	*b = (uint16_t *)buf;
2697 
2698 	while (len > 1) {
2699 		sum += *b++;
2700 		len -= 2;
2701 	}
2702 
2703 	if (len == 1)
2704 		sum += htons((*(unsigned char *)b) << 8);
2705 
2706 	return (sum);
2707 }
2708 
2709 /* ------------------------------------------------------------------------ */
2710 /* Function:    fr_calc_chksum						    */
2711 /* Returns:     void							    */
2712 /* Parameters:  fin - pointer to fr_info_t instance with packet data	    */
2713 /*              pkt - pointer to duplicated packet			    */
2714 /*                                                                          */
2715 /* Calculates all chksums (L3, L4) for packet pkt. Works for both IP	    */
2716 /* versions.								    */
2717 /* ------------------------------------------------------------------------ */
2718 void fr_calc_chksum(fin, pkt)
2719 fr_info_t *fin;
2720 mb_t *pkt;
2721 {
2722 	struct pseudo_hdr {
2723 		union {
2724 			struct in_addr	in4;
2725 #ifdef USE_INET6
2726 			struct in6_addr	in6;
2727 #endif
2728 		} src_addr;
2729 		union {
2730 			struct in_addr	in4;
2731 #ifdef USE_INET6
2732 			struct in6_addr	in6;
2733 #endif
2734 		} dst_addr;
2735 		char		zero;
2736 		char		proto;
2737 		uint16_t	len;
2738 	}	phdr;
2739 	uint32_t	sum, ip_sum;
2740 	void	*buf;
2741 	uint16_t	*l4_csum_p;
2742 	tcphdr_t	*tcp;
2743 	udphdr_t	*udp;
2744 	icmphdr_t	*icmp;
2745 #ifdef USE_INET6
2746 	struct icmp6_hdr	*icmp6;
2747 #endif
2748 	ip_t		*ip;
2749 	unsigned int	len;
2750 	int		pld_len;
2751 
2752 	/*
2753 	 * We need to pullup the packet to the single continuous buffer to avoid
2754 	 * potential misaligment of b_rptr member in mblk chain.
2755 	 */
2756 	if (pullupmsg(pkt, -1) == 0) {
2757 		cmn_err(CE_WARN, "Failed to pullup loopback pkt -> chksum"
2758 		    " will not be computed by IPF");
2759 		return;
2760 	}
2761 
2762 	/*
2763 	 * It is guaranteed IP header starts right at b_rptr, because we are
2764 	 * working with a copy of the original packet.
2765 	 *
2766 	 * Compute pseudo header chksum for TCP and UDP.
2767 	 */
2768 	if ((fin->fin_p == IPPROTO_UDP) ||
2769 	    (fin->fin_p == IPPROTO_TCP)) {
2770 		bzero(&phdr, sizeof (phdr));
2771 #ifdef USE_INET6
2772 		if (fin->fin_v == 6) {
2773 			phdr.src_addr.in6 = fin->fin_srcip6;
2774 			phdr.dst_addr.in6 = fin->fin_dstip6;
2775 		} else {
2776 			phdr.src_addr.in4 = fin->fin_src;
2777 			phdr.dst_addr.in4 = fin->fin_dst;
2778 		}
2779 #else
2780 		phdr.src_addr.in4 = fin->fin_src;
2781 		phdr.dst_addr.in4 = fin->fin_dst;
2782 #endif
2783 		phdr.zero = (char) 0;
2784 		phdr.proto = fin->fin_p;
2785 		phdr.len = htons((uint16_t)fin->fin_dlen);
2786 		sum = fr_buf_sum(&phdr, (unsigned int)sizeof (phdr));
2787 	} else {
2788 		sum = 0;
2789 	}
2790 
2791 	/*
2792 	 * Set pointer to the L4 chksum field in the packet, set buf pointer to
2793 	 * the L4 header start.
2794 	 */
2795 	switch (fin->fin_p) {
2796 		case IPPROTO_UDP:
2797 			udp = (udphdr_t *)(pkt->b_rptr + fin->fin_hlen);
2798 			l4_csum_p = &udp->uh_sum;
2799 			buf = udp;
2800 			break;
2801 		case IPPROTO_TCP:
2802 			tcp = (tcphdr_t *)(pkt->b_rptr + fin->fin_hlen);
2803 			l4_csum_p = &tcp->th_sum;
2804 			buf = tcp;
2805 			break;
2806 		case IPPROTO_ICMP:
2807 			icmp = (icmphdr_t *)(pkt->b_rptr + fin->fin_hlen);
2808 			l4_csum_p = &icmp->icmp_cksum;
2809 			buf = icmp;
2810 			break;
2811 #ifdef USE_INET6
2812 		case IPPROTO_ICMPV6:
2813 			icmp6 = (struct icmp6_hdr *)(pkt->b_rptr + fin->fin_hlen);
2814 			l4_csum_p = &icmp6->icmp6_cksum;
2815 			buf = icmp6;
2816 			break;
2817 #endif
2818 		default:
2819 			l4_csum_p = NULL;
2820 	}
2821 
2822 	/*
2823 	 * Compute L4 chksum if needed.
2824 	 */
2825 	if (l4_csum_p != NULL) {
2826 		*l4_csum_p = (uint16_t)0;
2827 		pld_len = fin->fin_dlen;
2828 		len = pkt->b_wptr - (unsigned char *)buf;
2829 		ASSERT(len == pld_len);
2830 		/*
2831 		 * Add payload sum to pseudoheader sum.
2832 		 */
2833 		sum += fr_buf_sum(buf, len);
2834 		while (sum >> 16)
2835 			sum = (sum & 0xFFFF) + (sum >> 16);
2836 
2837 		*l4_csum_p = ~((uint16_t)sum);
2838 		DTRACE_PROBE1(l4_sum, uint16_t, *l4_csum_p);
2839 	}
2840 
2841 	/*
2842 	 * The IP header chksum is needed just for IPv4.
2843 	 */
2844 	if (fin->fin_v == 4) {
2845 		/*
2846 		 * Compute IPv4 header chksum.
2847 		 */
2848 		ip = (ip_t *)pkt->b_rptr;
2849 		ip->ip_sum = (uint16_t)0;
2850 		ip_sum = fr_buf_sum(ip, (unsigned int)fin->fin_hlen);
2851 		while (ip_sum >> 16)
2852 			ip_sum = (ip_sum & 0xFFFF) + (ip_sum >> 16);
2853 
2854 		ip->ip_sum = ~((uint16_t)ip_sum);
2855 		DTRACE_PROBE1(l3_sum, uint16_t, ip->ip_sum);
2856 	}
2857 
2858 	return;
2859 }
2860 
2861 #endif	/* _KERNEL && SOLARIS2 >= 10 */
2862