if_pfsync.c revision 229770
1/*	$OpenBSD: if_pfsync.c,v 1.110 2009/02/24 05:39:19 dlg Exp $	*/
2
3/*
4 * Copyright (c) 2002 Michael Shalayeff
5 * All rights reserved.
6 *
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
9 * are met:
10 * 1. Redistributions of source code must retain the above copyright
11 *    notice, this list of conditions and the following disclaimer.
12 * 2. Redistributions in binary form must reproduce the above copyright
13 *    notice, this list of conditions and the following disclaimer in the
14 *    documentation and/or other materials provided with the distribution.
15 *
16 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
17 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
18 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
19 * IN NO EVENT SHALL THE AUTHOR OR HIS RELATIVES BE LIABLE FOR ANY DIRECT,
20 * INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
21 * (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
22 * SERVICES; LOSS OF MIND, USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
23 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT,
24 * STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING
25 * IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF
26 * THE POSSIBILITY OF SUCH DAMAGE.
27 */
28
29/*
30 * Copyright (c) 2009 David Gwynne <dlg@openbsd.org>
31 *
32 * Permission to use, copy, modify, and distribute this software for any
33 * purpose with or without fee is hereby granted, provided that the above
34 * copyright notice and this permission notice appear in all copies.
35 *
36 * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
37 * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
38 * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
39 * ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
40 * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN
41 * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
42 * OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
43 */
44
45/*
46 * Revisions picked from OpenBSD after revision 1.110 import:
47 * 1.118, 1.124, 1.148, 1.149, 1.151, 1.171 - fixes to bulk updates
48 * 1.120, 1.175 - use monotonic time_uptime
49 * 1.122 - reduce number of updates for non-TCP sessions
50 */
51
52#ifdef __FreeBSD__
53#include "opt_inet.h"
54#include "opt_inet6.h"
55#include "opt_pf.h"
56
57#include <sys/cdefs.h>
58__FBSDID("$FreeBSD: stable/9/sys/contrib/pf/net/if_pfsync.c 229770 2012-01-07 11:01:35Z glebius $");
59
60#define	NBPFILTER	1
61
62#ifdef DEV_PFSYNC
63#define	NPFSYNC		DEV_PFSYNC
64#else
65#define	NPFSYNC		0
66#endif
67
68#ifdef DEV_CARP
69#define	NCARP		DEV_CARP
70#else
71#define	NCARP		0
72#endif
73#endif /* __FreeBSD__ */
74
75#include <sys/param.h>
76#include <sys/kernel.h>
77#ifdef __FreeBSD__
78#include <sys/bus.h>
79#include <sys/interrupt.h>
80#include <sys/priv.h>
81#endif
82#include <sys/proc.h>
83#include <sys/systm.h>
84#include <sys/time.h>
85#include <sys/mbuf.h>
86#include <sys/socket.h>
87#ifdef __FreeBSD__
88#include <sys/endian.h>
89#include <sys/malloc.h>
90#include <sys/module.h>
91#include <sys/sockio.h>
92#include <sys/taskqueue.h>
93#include <sys/lock.h>
94#include <sys/mutex.h>
95#else
96#include <sys/ioctl.h>
97#include <sys/timeout.h>
98#endif
99#include <sys/sysctl.h>
100#ifndef __FreeBSD__
101#include <sys/pool.h>
102#endif
103
104#include <net/if.h>
105#ifdef __FreeBSD__
106#include <net/if_clone.h>
107#endif
108#include <net/if_types.h>
109#include <net/route.h>
110#include <net/bpf.h>
111#include <net/netisr.h>
112#ifdef __FreeBSD__
113#include <net/vnet.h>
114#endif
115
116#include <netinet/in.h>
117#include <netinet/if_ether.h>
118#include <netinet/tcp.h>
119#include <netinet/tcp_seq.h>
120
121#ifdef	INET
122#include <netinet/in_systm.h>
123#include <netinet/in_var.h>
124#include <netinet/ip.h>
125#include <netinet/ip_var.h>
126#endif
127
128#ifdef INET6
129#include <netinet6/nd6.h>
130#endif /* INET6 */
131
132#ifndef __FreeBSD__
133#include "carp.h"
134#endif
135#if NCARP > 0
136#include <netinet/ip_carp.h>
137#endif
138
139#include <net/pfvar.h>
140#include <net/if_pfsync.h>
141
142#ifndef __FreeBSD__
143#include "bpfilter.h"
144#include "pfsync.h"
145#endif
146
147#define PFSYNC_MINPKT ( \
148	sizeof(struct ip) + \
149	sizeof(struct pfsync_header) + \
150	sizeof(struct pfsync_subheader) + \
151	sizeof(struct pfsync_eof))
152
153struct pfsync_pkt {
154	struct ip *ip;
155	struct in_addr src;
156	u_int8_t flags;
157};
158
159int	pfsync_input_hmac(struct mbuf *, int);
160
161int	pfsync_upd_tcp(struct pf_state *, struct pfsync_state_peer *,
162	    struct pfsync_state_peer *);
163
164int	pfsync_in_clr(struct pfsync_pkt *, struct mbuf *, int, int);
165int	pfsync_in_ins(struct pfsync_pkt *, struct mbuf *, int, int);
166int	pfsync_in_iack(struct pfsync_pkt *, struct mbuf *, int, int);
167int	pfsync_in_upd(struct pfsync_pkt *, struct mbuf *, int, int);
168int	pfsync_in_upd_c(struct pfsync_pkt *, struct mbuf *, int, int);
169int	pfsync_in_ureq(struct pfsync_pkt *, struct mbuf *, int, int);
170int	pfsync_in_del(struct pfsync_pkt *, struct mbuf *, int, int);
171int	pfsync_in_del_c(struct pfsync_pkt *, struct mbuf *, int, int);
172int	pfsync_in_bus(struct pfsync_pkt *, struct mbuf *, int, int);
173int	pfsync_in_tdb(struct pfsync_pkt *, struct mbuf *, int, int);
174int	pfsync_in_eof(struct pfsync_pkt *, struct mbuf *, int, int);
175
176int	pfsync_in_error(struct pfsync_pkt *, struct mbuf *, int, int);
177
178int	(*pfsync_acts[])(struct pfsync_pkt *, struct mbuf *, int, int) = {
179	pfsync_in_clr,			/* PFSYNC_ACT_CLR */
180	pfsync_in_ins,			/* PFSYNC_ACT_INS */
181	pfsync_in_iack,			/* PFSYNC_ACT_INS_ACK */
182	pfsync_in_upd,			/* PFSYNC_ACT_UPD */
183	pfsync_in_upd_c,		/* PFSYNC_ACT_UPD_C */
184	pfsync_in_ureq,			/* PFSYNC_ACT_UPD_REQ */
185	pfsync_in_del,			/* PFSYNC_ACT_DEL */
186	pfsync_in_del_c,		/* PFSYNC_ACT_DEL_C */
187	pfsync_in_error,		/* PFSYNC_ACT_INS_F */
188	pfsync_in_error,		/* PFSYNC_ACT_DEL_F */
189	pfsync_in_bus,			/* PFSYNC_ACT_BUS */
190	pfsync_in_tdb,			/* PFSYNC_ACT_TDB */
191	pfsync_in_eof			/* PFSYNC_ACT_EOF */
192};
193
194struct pfsync_q {
195	int		(*write)(struct pf_state *, struct mbuf *, int);
196	size_t		len;
197	u_int8_t	action;
198};
199
200/* we have one of these for every PFSYNC_S_ */
201int	pfsync_out_state(struct pf_state *, struct mbuf *, int);
202int	pfsync_out_iack(struct pf_state *, struct mbuf *, int);
203int	pfsync_out_upd_c(struct pf_state *, struct mbuf *, int);
204int	pfsync_out_del(struct pf_state *, struct mbuf *, int);
205
206struct pfsync_q pfsync_qs[] = {
207	{ pfsync_out_state, sizeof(struct pfsync_state),   PFSYNC_ACT_INS },
208	{ pfsync_out_iack,  sizeof(struct pfsync_ins_ack), PFSYNC_ACT_INS_ACK },
209	{ pfsync_out_state, sizeof(struct pfsync_state),   PFSYNC_ACT_UPD },
210	{ pfsync_out_upd_c, sizeof(struct pfsync_upd_c),   PFSYNC_ACT_UPD_C },
211	{ pfsync_out_del,   sizeof(struct pfsync_del_c),   PFSYNC_ACT_DEL_C }
212};
213
214void	pfsync_q_ins(struct pf_state *, int);
215void	pfsync_q_del(struct pf_state *);
216
217struct pfsync_upd_req_item {
218	TAILQ_ENTRY(pfsync_upd_req_item)	ur_entry;
219	struct pfsync_upd_req			ur_msg;
220};
221TAILQ_HEAD(pfsync_upd_reqs, pfsync_upd_req_item);
222
223struct pfsync_deferral {
224	TAILQ_ENTRY(pfsync_deferral)		 pd_entry;
225	struct pf_state				*pd_st;
226	struct mbuf				*pd_m;
227#ifdef __FreeBSD__
228	struct callout				 pd_tmo;
229#else
230	struct timeout				 pd_tmo;
231#endif
232};
233TAILQ_HEAD(pfsync_deferrals, pfsync_deferral);
234
235#define PFSYNC_PLSIZE	MAX(sizeof(struct pfsync_upd_req_item), \
236			    sizeof(struct pfsync_deferral))
237
238#ifdef notyet
239int	pfsync_out_tdb(struct tdb *, struct mbuf *, int);
240#endif
241
242struct pfsync_softc {
243#ifdef __FreeBSD__
244	struct ifnet		*sc_ifp;
245#else
246	struct ifnet		 sc_if;
247#endif
248	struct ifnet		*sc_sync_if;
249
250#ifdef __FreeBSD__
251	uma_zone_t		 sc_pool;
252#else
253	struct pool		 sc_pool;
254#endif
255
256	struct ip_moptions	 sc_imo;
257
258	struct in_addr		 sc_sync_peer;
259	u_int8_t		 sc_maxupdates;
260#ifdef __FreeBSD__
261	int			 pfsync_sync_ok;
262#endif
263
264	struct ip		 sc_template;
265
266	struct pf_state_queue	 sc_qs[PFSYNC_S_COUNT];
267	size_t			 sc_len;
268
269	struct pfsync_upd_reqs	 sc_upd_req_list;
270
271	struct pfsync_deferrals	 sc_deferrals;
272	u_int			 sc_deferred;
273
274	void			*sc_plus;
275	size_t			 sc_pluslen;
276
277	u_int32_t		 sc_ureq_sent;
278	int			 sc_bulk_tries;
279#ifdef __FreeBSD__
280	struct callout		 sc_bulkfail_tmo;
281#else
282	struct timeout		 sc_bulkfail_tmo;
283#endif
284
285	u_int32_t		 sc_ureq_received;
286	struct pf_state		*sc_bulk_next;
287	struct pf_state		*sc_bulk_last;
288#ifdef __FreeBSD__
289	struct callout		 sc_bulk_tmo;
290#else
291	struct timeout		 sc_bulk_tmo;
292#endif
293
294	TAILQ_HEAD(, tdb)	 sc_tdb_q;
295
296#ifdef __FreeBSD__
297	struct callout		 sc_tmo;
298#else
299	struct timeout		 sc_tmo;
300#endif
301#ifdef __FreeBSD__
302	eventhandler_tag	 sc_detachtag;
303#endif
304
305};
306
307#ifdef __FreeBSD__
308static VNET_DEFINE(struct pfsync_softc	*, pfsyncif) = NULL;
309#define	V_pfsyncif		VNET(pfsyncif)
310
311static VNET_DEFINE(struct pfsyncstats, pfsyncstats);
312#define	V_pfsyncstats		VNET(pfsyncstats)
313
314SYSCTL_NODE(_net, OID_AUTO, pfsync, CTLFLAG_RW, 0, "PFSYNC");
315SYSCTL_VNET_STRUCT(_net_pfsync, OID_AUTO, stats, CTLFLAG_RW,
316    &VNET_NAME(pfsyncstats), pfsyncstats,
317    "PFSYNC statistics (struct pfsyncstats, net/if_pfsync.h)");
318#else
319struct pfsync_softc	*pfsyncif = NULL;
320struct pfsyncstats	 pfsyncstats;
321#define	V_pfsyncstats	 pfsyncstats
322#endif
323
324#ifdef __FreeBSD__
325static void	pfsyncintr(void *);
326struct pfsync_swi {
327	void *	pfsync_swi_cookie;
328};
329static struct pfsync_swi	 pfsync_swi;
330#define	schednetisr(p)	swi_sched(pfsync_swi.pfsync_swi_cookie, 0)
331#define	NETISR_PFSYNC
332#endif
333
334void	pfsyncattach(int);
335#ifdef __FreeBSD__
336int	pfsync_clone_create(struct if_clone *, int, caddr_t);
337void	pfsync_clone_destroy(struct ifnet *);
338#else
339int	pfsync_clone_create(struct if_clone *, int);
340int	pfsync_clone_destroy(struct ifnet *);
341#endif
342int	pfsync_alloc_scrub_memory(struct pfsync_state_peer *,
343	    struct pf_state_peer *);
344void	pfsync_update_net_tdb(struct pfsync_tdb *);
345int	pfsyncoutput(struct ifnet *, struct mbuf *, struct sockaddr *,
346#ifdef __FreeBSD__
347	    struct route *);
348#else
349	    struct rtentry *);
350#endif
351int	pfsyncioctl(struct ifnet *, u_long, caddr_t);
352void	pfsyncstart(struct ifnet *);
353
354struct mbuf *pfsync_if_dequeue(struct ifnet *);
355struct mbuf *pfsync_get_mbuf(struct pfsync_softc *);
356
357void	pfsync_deferred(struct pf_state *, int);
358void	pfsync_undefer(struct pfsync_deferral *, int);
359void	pfsync_defer_tmo(void *);
360
361void	pfsync_request_update(u_int32_t, u_int64_t);
362void	pfsync_update_state_req(struct pf_state *);
363
364void	pfsync_drop(struct pfsync_softc *);
365void	pfsync_sendout(void);
366void	pfsync_send_plus(void *, size_t);
367int	pfsync_tdb_sendout(struct pfsync_softc *);
368int	pfsync_sendout_mbuf(struct pfsync_softc *, struct mbuf *);
369void	pfsync_timeout(void *);
370void	pfsync_tdb_timeout(void *);
371void	pfsync_send_bus(struct pfsync_softc *, u_int8_t);
372
373void	pfsync_bulk_start(void);
374void	pfsync_bulk_status(u_int8_t);
375void	pfsync_bulk_update(void *);
376void	pfsync_bulk_fail(void *);
377
378#ifdef __FreeBSD__
379void	pfsync_ifdetach(void *, struct ifnet *);
380
381/* XXX: ugly */
382#define	betoh64		(unsigned long long)be64toh
383#define	timeout_del	callout_stop
384#endif
385
386#define PFSYNC_MAX_BULKTRIES	12
387#ifndef __FreeBSD__
388int	pfsync_sync_ok;
389#endif
390
391#ifdef __FreeBSD__
392IFC_SIMPLE_DECLARE(pfsync, 1);
393#else
394struct if_clone	pfsync_cloner =
395    IF_CLONE_INITIALIZER("pfsync", pfsync_clone_create, pfsync_clone_destroy);
396#endif
397
398void
399pfsyncattach(int npfsync)
400{
401	if_clone_attach(&pfsync_cloner);
402}
403int
404#ifdef __FreeBSD__
405pfsync_clone_create(struct if_clone *ifc, int unit, caddr_t param)
406#else
407pfsync_clone_create(struct if_clone *ifc, int unit)
408#endif
409{
410	struct pfsync_softc *sc;
411	struct ifnet *ifp;
412	int q;
413
414	if (unit != 0)
415		return (EINVAL);
416
417#ifndef __FreeBSD__
418	pfsync_sync_ok = 1;
419#endif
420
421	sc = malloc(sizeof(struct pfsync_softc), M_DEVBUF, M_NOWAIT | M_ZERO);
422	if (sc == NULL)
423		return (ENOMEM);
424
425	for (q = 0; q < PFSYNC_S_COUNT; q++)
426		TAILQ_INIT(&sc->sc_qs[q]);
427
428#ifdef __FreeBSD__
429	sc->pfsync_sync_ok = 1;
430	sc->sc_pool = uma_zcreate("pfsync", PFSYNC_PLSIZE,
431			NULL, NULL, NULL, NULL, UMA_ALIGN_PTR, 0);
432	if (sc->sc_pool == NULL) {
433		free(sc, M_DEVBUF);
434		return (ENOMEM);
435	}
436#else
437	pool_init(&sc->sc_pool, PFSYNC_PLSIZE, 0, 0, 0, "pfsync", NULL);
438#endif
439	TAILQ_INIT(&sc->sc_upd_req_list);
440	TAILQ_INIT(&sc->sc_deferrals);
441	sc->sc_deferred = 0;
442
443	TAILQ_INIT(&sc->sc_tdb_q);
444
445	sc->sc_len = PFSYNC_MINPKT;
446	sc->sc_maxupdates = 128;
447
448#ifdef __FreeBSD__
449	sc->sc_imo.imo_membership = (struct in_multi **)malloc(
450	    (sizeof(struct in_multi *) * IP_MIN_MEMBERSHIPS), M_DEVBUF,
451	    M_NOWAIT | M_ZERO);
452	sc->sc_imo.imo_max_memberships = IP_MIN_MEMBERSHIPS;
453	sc->sc_imo.imo_multicast_vif = -1;
454#else
455	sc->sc_imo.imo_membership = (struct in_multi **)malloc(
456	    (sizeof(struct in_multi *) * IP_MIN_MEMBERSHIPS), M_IPMOPTS,
457	    M_WAITOK | M_ZERO);
458	sc->sc_imo.imo_max_memberships = IP_MIN_MEMBERSHIPS;
459#endif
460
461#ifdef __FreeBSD__
462	ifp = sc->sc_ifp = if_alloc(IFT_PFSYNC);
463	if (ifp == NULL) {
464		free(sc->sc_imo.imo_membership, M_DEVBUF);
465		uma_zdestroy(sc->sc_pool);
466		free(sc, M_DEVBUF);
467		return (ENOSPC);
468	}
469	if_initname(ifp, ifc->ifc_name, unit);
470
471	sc->sc_detachtag = EVENTHANDLER_REGISTER(ifnet_departure_event,
472#ifdef __FreeBSD__
473	    pfsync_ifdetach, V_pfsyncif, EVENTHANDLER_PRI_ANY);
474#else
475	    pfsync_ifdetach, pfsyncif, EVENTHANDLER_PRI_ANY);
476#endif
477	if (sc->sc_detachtag == NULL) {
478		if_free(ifp);
479		free(sc->sc_imo.imo_membership, M_DEVBUF);
480		uma_zdestroy(sc->sc_pool);
481		free(sc, M_DEVBUF);
482		return (ENOSPC);
483	}
484#else
485	ifp = &sc->sc_if;
486	snprintf(ifp->if_xname, sizeof ifp->if_xname, "pfsync%d", unit);
487#endif
488	ifp->if_softc = sc;
489	ifp->if_ioctl = pfsyncioctl;
490	ifp->if_output = pfsyncoutput;
491	ifp->if_start = pfsyncstart;
492	ifp->if_type = IFT_PFSYNC;
493	ifp->if_snd.ifq_maxlen = ifqmaxlen;
494	ifp->if_hdrlen = sizeof(struct pfsync_header);
495	ifp->if_mtu = 1500; /* XXX */
496#ifdef __FreeBSD__
497	callout_init(&sc->sc_tmo, CALLOUT_MPSAFE);
498	callout_init_mtx(&sc->sc_bulk_tmo, &pf_task_mtx, 0);
499	callout_init(&sc->sc_bulkfail_tmo, CALLOUT_MPSAFE);
500#else
501	ifp->if_hardmtu = MCLBYTES; /* XXX */
502	timeout_set(&sc->sc_tmo, pfsync_timeout, sc);
503	timeout_set(&sc->sc_bulk_tmo, pfsync_bulk_update, sc);
504	timeout_set(&sc->sc_bulkfail_tmo, pfsync_bulk_fail, sc);
505#endif
506
507	if_attach(ifp);
508#ifndef __FreeBSD__
509	if_alloc_sadl(ifp);
510#endif
511
512#if NCARP > 0
513	if_addgroup(ifp, "carp");
514#endif
515
516#if NBPFILTER > 0
517#ifdef __FreeBSD__
518	bpfattach(ifp, DLT_PFSYNC, PFSYNC_HDRLEN);
519#else
520	bpfattach(&sc->sc_if.if_bpf, ifp, DLT_PFSYNC, PFSYNC_HDRLEN);
521#endif
522#endif
523
524#ifdef __FreeBSD__
525	V_pfsyncif = sc;
526#else
527	pfsyncif = sc;
528#endif
529
530	return (0);
531}
532
533#ifdef __FreeBSD__
534void
535#else
536int
537#endif
538pfsync_clone_destroy(struct ifnet *ifp)
539{
540	struct pfsync_softc *sc = ifp->if_softc;
541
542#ifdef __FreeBSD__
543	EVENTHANDLER_DEREGISTER(ifnet_departure_event, sc->sc_detachtag);
544	PF_LOCK();
545#endif
546	timeout_del(&sc->sc_bulkfail_tmo);
547	timeout_del(&sc->sc_bulk_tmo);
548	timeout_del(&sc->sc_tmo);
549#ifdef __FreeBSD__
550	PF_UNLOCK();
551#endif
552#if NCARP > 0
553#ifdef notyet
554#ifdef __FreeBSD__
555	if (!sc->pfsync_sync_ok)
556#else
557	if (!pfsync_sync_ok)
558#endif
559		carp_group_demote_adj(&sc->sc_if, -1);
560#endif
561#endif
562#if NBPFILTER > 0
563	bpfdetach(ifp);
564#endif
565	if_detach(ifp);
566
567	pfsync_drop(sc);
568
569	while (sc->sc_deferred > 0)
570		pfsync_undefer(TAILQ_FIRST(&sc->sc_deferrals), 0);
571
572#ifdef __FreeBSD__
573	UMA_DESTROY(sc->sc_pool);
574#else
575	pool_destroy(&sc->sc_pool);
576#endif
577#ifdef __FreeBSD__
578	if_free(ifp);
579	free(sc->sc_imo.imo_membership, M_DEVBUF);
580#else
581	free(sc->sc_imo.imo_membership, M_IPMOPTS);
582#endif
583	free(sc, M_DEVBUF);
584
585#ifdef __FreeBSD__
586	V_pfsyncif = NULL;
587#else
588	pfsyncif = NULL;
589#endif
590
591#ifndef __FreeBSD__
592	return (0);
593#endif
594}
595
596struct mbuf *
597pfsync_if_dequeue(struct ifnet *ifp)
598{
599	struct mbuf *m;
600#ifndef __FreeBSD__
601	int s;
602#endif
603
604#ifdef __FreeBSD__
605	IF_LOCK(&ifp->if_snd);
606	_IF_DROP(&ifp->if_snd);
607	_IF_DEQUEUE(&ifp->if_snd, m);
608	IF_UNLOCK(&ifp->if_snd);
609#else
610	s = splnet();
611	IF_DEQUEUE(&ifp->if_snd, m);
612	splx(s);
613#endif
614
615	return (m);
616}
617
618/*
619 * Start output on the pfsync interface.
620 */
621void
622pfsyncstart(struct ifnet *ifp)
623{
624	struct mbuf *m;
625
626	while ((m = pfsync_if_dequeue(ifp)) != NULL) {
627#ifndef __FreeBSD__
628		IF_DROP(&ifp->if_snd);
629#endif
630		m_freem(m);
631	}
632}
633
634int
635pfsync_alloc_scrub_memory(struct pfsync_state_peer *s,
636    struct pf_state_peer *d)
637{
638	if (s->scrub.scrub_flag && d->scrub == NULL) {
639#ifdef __FreeBSD__
640		d->scrub = pool_get(&V_pf_state_scrub_pl, PR_NOWAIT | PR_ZERO);
641#else
642		d->scrub = pool_get(&pf_state_scrub_pl, PR_NOWAIT | PR_ZERO);
643#endif
644		if (d->scrub == NULL)
645			return (ENOMEM);
646	}
647
648	return (0);
649}
650
651#ifndef __FreeBSD__
652void
653pfsync_state_export(struct pfsync_state *sp, struct pf_state *st)
654{
655	bzero(sp, sizeof(struct pfsync_state));
656
657	/* copy from state key */
658	sp->key[PF_SK_WIRE].addr[0] = st->key[PF_SK_WIRE]->addr[0];
659	sp->key[PF_SK_WIRE].addr[1] = st->key[PF_SK_WIRE]->addr[1];
660	sp->key[PF_SK_WIRE].port[0] = st->key[PF_SK_WIRE]->port[0];
661	sp->key[PF_SK_WIRE].port[1] = st->key[PF_SK_WIRE]->port[1];
662	sp->key[PF_SK_STACK].addr[0] = st->key[PF_SK_STACK]->addr[0];
663	sp->key[PF_SK_STACK].addr[1] = st->key[PF_SK_STACK]->addr[1];
664	sp->key[PF_SK_STACK].port[0] = st->key[PF_SK_STACK]->port[0];
665	sp->key[PF_SK_STACK].port[1] = st->key[PF_SK_STACK]->port[1];
666	sp->proto = st->key[PF_SK_WIRE]->proto;
667	sp->af = st->key[PF_SK_WIRE]->af;
668
669	/* copy from state */
670	strlcpy(sp->ifname, st->kif->pfik_name, sizeof(sp->ifname));
671	bcopy(&st->rt_addr, &sp->rt_addr, sizeof(sp->rt_addr));
672	sp->creation = htonl(time_uptime - st->creation);
673	sp->expire = pf_state_expires(st);
674	if (sp->expire <= time_second)
675		sp->expire = htonl(0);
676	else
677		sp->expire = htonl(sp->expire - time_second);
678
679	sp->direction = st->direction;
680	sp->log = st->log;
681	sp->timeout = st->timeout;
682	sp->state_flags = st->state_flags;
683	if (st->src_node)
684		sp->sync_flags |= PFSYNC_FLAG_SRCNODE;
685	if (st->nat_src_node)
686		sp->sync_flags |= PFSYNC_FLAG_NATSRCNODE;
687
688	bcopy(&st->id, &sp->id, sizeof(sp->id));
689	sp->creatorid = st->creatorid;
690	pf_state_peer_hton(&st->src, &sp->src);
691	pf_state_peer_hton(&st->dst, &sp->dst);
692
693	if (st->rule.ptr == NULL)
694		sp->rule = htonl(-1);
695	else
696		sp->rule = htonl(st->rule.ptr->nr);
697	if (st->anchor.ptr == NULL)
698		sp->anchor = htonl(-1);
699	else
700		sp->anchor = htonl(st->anchor.ptr->nr);
701	if (st->nat_rule.ptr == NULL)
702		sp->nat_rule = htonl(-1);
703	else
704		sp->nat_rule = htonl(st->nat_rule.ptr->nr);
705
706	pf_state_counter_hton(st->packets[0], sp->packets[0]);
707	pf_state_counter_hton(st->packets[1], sp->packets[1]);
708	pf_state_counter_hton(st->bytes[0], sp->bytes[0]);
709	pf_state_counter_hton(st->bytes[1], sp->bytes[1]);
710
711}
712#endif
713
714int
715pfsync_state_import(struct pfsync_state *sp, u_int8_t flags)
716{
717	struct pf_state	*st = NULL;
718	struct pf_state_key *skw = NULL, *sks = NULL;
719	struct pf_rule *r = NULL;
720	struct pfi_kif	*kif;
721	int pool_flags;
722	int error;
723
724	PF_LOCK_ASSERT();
725
726#ifdef __FreeBSD__
727	if (sp->creatorid == 0 && V_pf_status.debug >= PF_DEBUG_MISC) {
728#else
729	if (sp->creatorid == 0 && pf_status.debug >= PF_DEBUG_MISC) {
730#endif
731		printf("pfsync_state_import: invalid creator id:"
732		    " %08x\n", ntohl(sp->creatorid));
733		return (EINVAL);
734	}
735
736	if ((kif = pfi_kif_get(sp->ifname)) == NULL) {
737#ifdef __FreeBSD__
738		if (V_pf_status.debug >= PF_DEBUG_MISC)
739#else
740		if (pf_status.debug >= PF_DEBUG_MISC)
741#endif
742			printf("pfsync_state_import: "
743			    "unknown interface: %s\n", sp->ifname);
744		if (flags & PFSYNC_SI_IOCTL)
745			return (EINVAL);
746		return (0);	/* skip this state */
747	}
748
749	/*
750	 * If the ruleset checksums match or the state is coming from the ioctl,
751	 * it's safe to associate the state with the rule of that number.
752	 */
753	if (sp->rule != htonl(-1) && sp->anchor == htonl(-1) &&
754	    (flags & (PFSYNC_SI_IOCTL | PFSYNC_SI_CKSUM)) && ntohl(sp->rule) <
755	    pf_main_ruleset.rules[PF_RULESET_FILTER].active.rcount)
756		r = pf_main_ruleset.rules[
757		    PF_RULESET_FILTER].active.ptr_array[ntohl(sp->rule)];
758	else
759#ifdef __FreeBSD__
760		r = &V_pf_default_rule;
761#else
762		r = &pf_default_rule;
763#endif
764
765	if ((r->max_states && r->states_cur >= r->max_states))
766		goto cleanup;
767
768#ifdef __FreeBSD__
769	if (flags & PFSYNC_SI_IOCTL)
770		pool_flags = PR_WAITOK | PR_ZERO;
771	else
772		pool_flags = PR_NOWAIT | PR_ZERO;
773
774	if ((st = pool_get(&V_pf_state_pl, pool_flags)) == NULL)
775		goto cleanup;
776#else
777	if (flags & PFSYNC_SI_IOCTL)
778		pool_flags = PR_WAITOK | PR_LIMITFAIL | PR_ZERO;
779	else
780		pool_flags = PR_LIMITFAIL | PR_ZERO;
781
782	if ((st = pool_get(&pf_state_pl, pool_flags)) == NULL)
783		goto cleanup;
784#endif
785
786	if ((skw = pf_alloc_state_key(pool_flags)) == NULL)
787		goto cleanup;
788
789	if (PF_ANEQ(&sp->key[PF_SK_WIRE].addr[0],
790	    &sp->key[PF_SK_STACK].addr[0], sp->af) ||
791	    PF_ANEQ(&sp->key[PF_SK_WIRE].addr[1],
792	    &sp->key[PF_SK_STACK].addr[1], sp->af) ||
793	    sp->key[PF_SK_WIRE].port[0] != sp->key[PF_SK_STACK].port[0] ||
794	    sp->key[PF_SK_WIRE].port[1] != sp->key[PF_SK_STACK].port[1]) {
795		if ((sks = pf_alloc_state_key(pool_flags)) == NULL)
796			goto cleanup;
797	} else
798		sks = skw;
799
800	/* allocate memory for scrub info */
801	if (pfsync_alloc_scrub_memory(&sp->src, &st->src) ||
802	    pfsync_alloc_scrub_memory(&sp->dst, &st->dst))
803		goto cleanup;
804
805	/* copy to state key(s) */
806	skw->addr[0] = sp->key[PF_SK_WIRE].addr[0];
807	skw->addr[1] = sp->key[PF_SK_WIRE].addr[1];
808	skw->port[0] = sp->key[PF_SK_WIRE].port[0];
809	skw->port[1] = sp->key[PF_SK_WIRE].port[1];
810	skw->proto = sp->proto;
811	skw->af = sp->af;
812	if (sks != skw) {
813		sks->addr[0] = sp->key[PF_SK_STACK].addr[0];
814		sks->addr[1] = sp->key[PF_SK_STACK].addr[1];
815		sks->port[0] = sp->key[PF_SK_STACK].port[0];
816		sks->port[1] = sp->key[PF_SK_STACK].port[1];
817		sks->proto = sp->proto;
818		sks->af = sp->af;
819	}
820
821	/* copy to state */
822	bcopy(&sp->rt_addr, &st->rt_addr, sizeof(st->rt_addr));
823	st->creation = time_uptime - ntohl(sp->creation);
824	st->expire = time_second;
825	if (sp->expire) {
826		/* XXX No adaptive scaling. */
827		st->expire -= r->timeout[sp->timeout] - ntohl(sp->expire);
828	}
829
830	st->expire = ntohl(sp->expire) + time_second;
831	st->direction = sp->direction;
832	st->log = sp->log;
833	st->timeout = sp->timeout;
834	st->state_flags = sp->state_flags;
835
836	bcopy(sp->id, &st->id, sizeof(st->id));
837	st->creatorid = sp->creatorid;
838	pf_state_peer_ntoh(&sp->src, &st->src);
839	pf_state_peer_ntoh(&sp->dst, &st->dst);
840
841	st->rule.ptr = r;
842	st->nat_rule.ptr = NULL;
843	st->anchor.ptr = NULL;
844	st->rt_kif = NULL;
845
846	st->pfsync_time = time_uptime;
847	st->sync_state = PFSYNC_S_NONE;
848
849	/* XXX when we have nat_rule/anchors, use STATE_INC_COUNTERS */
850	r->states_cur++;
851	r->states_tot++;
852
853	if (!ISSET(flags, PFSYNC_SI_IOCTL))
854		SET(st->state_flags, PFSTATE_NOSYNC);
855
856	if ((error = pf_state_insert(kif, skw, sks, st)) != 0) {
857		/* XXX when we have nat_rule/anchors, use STATE_DEC_COUNTERS */
858		r->states_cur--;
859		goto cleanup_state;
860	}
861
862	if (!ISSET(flags, PFSYNC_SI_IOCTL)) {
863		CLR(st->state_flags, PFSTATE_NOSYNC);
864		if (ISSET(st->state_flags, PFSTATE_ACK)) {
865			pfsync_q_ins(st, PFSYNC_S_IACK);
866#ifdef __FreeBSD__
867			pfsync_sendout();
868#else
869			schednetisr(NETISR_PFSYNC);
870#endif
871		}
872	}
873	CLR(st->state_flags, PFSTATE_ACK);
874
875	return (0);
876
877cleanup:
878	error = ENOMEM;
879	if (skw == sks)
880		sks = NULL;
881#ifdef __FreeBSD__
882	if (skw != NULL)
883		pool_put(&V_pf_state_key_pl, skw);
884	if (sks != NULL)
885		pool_put(&V_pf_state_key_pl, sks);
886#else
887	if (skw != NULL)
888		pool_put(&pf_state_key_pl, skw);
889	if (sks != NULL)
890		pool_put(&pf_state_key_pl, sks);
891#endif
892
893cleanup_state:	/* pf_state_insert frees the state keys */
894	if (st) {
895#ifdef __FreeBSD__
896		if (st->dst.scrub)
897			pool_put(&V_pf_state_scrub_pl, st->dst.scrub);
898		if (st->src.scrub)
899			pool_put(&V_pf_state_scrub_pl, st->src.scrub);
900		pool_put(&V_pf_state_pl, st);
901#else
902		if (st->dst.scrub)
903			pool_put(&pf_state_scrub_pl, st->dst.scrub);
904		if (st->src.scrub)
905			pool_put(&pf_state_scrub_pl, st->src.scrub);
906		pool_put(&pf_state_pl, st);
907#endif
908	}
909	return (error);
910}
911
912void
913#ifdef __FreeBSD__
914pfsync_input(struct mbuf *m, __unused int off)
915#else
916pfsync_input(struct mbuf *m, ...)
917#endif
918{
919#ifdef __FreeBSD__
920	struct pfsync_softc *sc = V_pfsyncif;
921#else
922	struct pfsync_softc *sc = pfsyncif;
923#endif
924	struct pfsync_pkt pkt;
925	struct ip *ip = mtod(m, struct ip *);
926	struct pfsync_header *ph;
927	struct pfsync_subheader subh;
928
929	int offset;
930	int rv;
931
932	V_pfsyncstats.pfsyncs_ipackets++;
933
934	/* verify that we have a sync interface configured */
935#ifdef __FreeBSD__
936	if (!sc || !sc->sc_sync_if || !V_pf_status.running)
937#else
938	if (!sc || !sc->sc_sync_if || !pf_status.running)
939#endif
940		goto done;
941
942	/* verify that the packet came in on the right interface */
943	if (sc->sc_sync_if != m->m_pkthdr.rcvif) {
944		V_pfsyncstats.pfsyncs_badif++;
945		goto done;
946	}
947
948#ifdef __FreeBSD__
949	sc->sc_ifp->if_ipackets++;
950	sc->sc_ifp->if_ibytes += m->m_pkthdr.len;
951#else
952	sc->sc_if.if_ipackets++;
953	sc->sc_if.if_ibytes += m->m_pkthdr.len;
954#endif
955	/* verify that the IP TTL is 255. */
956	if (ip->ip_ttl != PFSYNC_DFLTTL) {
957		V_pfsyncstats.pfsyncs_badttl++;
958		goto done;
959	}
960
961	offset = ip->ip_hl << 2;
962	if (m->m_pkthdr.len < offset + sizeof(*ph)) {
963		V_pfsyncstats.pfsyncs_hdrops++;
964		goto done;
965	}
966
967	if (offset + sizeof(*ph) > m->m_len) {
968		if (m_pullup(m, offset + sizeof(*ph)) == NULL) {
969			V_pfsyncstats.pfsyncs_hdrops++;
970			return;
971		}
972		ip = mtod(m, struct ip *);
973	}
974	ph = (struct pfsync_header *)((char *)ip + offset);
975
976	/* verify the version */
977	if (ph->version != PFSYNC_VERSION) {
978		V_pfsyncstats.pfsyncs_badver++;
979		goto done;
980	}
981
982#if 0
983	if (pfsync_input_hmac(m, offset) != 0) {
984		/* XXX stats */
985		goto done;
986	}
987#endif
988
989	/* Cheaper to grab this now than having to mess with mbufs later */
990	pkt.ip = ip;
991	pkt.src = ip->ip_src;
992	pkt.flags = 0;
993
994#ifdef __FreeBSD__
995	if (!bcmp(&ph->pfcksum, &V_pf_status.pf_chksum, PF_MD5_DIGEST_LENGTH))
996#else
997	if (!bcmp(&ph->pfcksum, &pf_status.pf_chksum, PF_MD5_DIGEST_LENGTH))
998#endif
999		pkt.flags |= PFSYNC_SI_CKSUM;
1000
1001	offset += sizeof(*ph);
1002	for (;;) {
1003		m_copydata(m, offset, sizeof(subh), (caddr_t)&subh);
1004		offset += sizeof(subh);
1005
1006		if (subh.action >= PFSYNC_ACT_MAX) {
1007			V_pfsyncstats.pfsyncs_badact++;
1008			goto done;
1009		}
1010
1011		rv = (*pfsync_acts[subh.action])(&pkt, m, offset,
1012		    ntohs(subh.count));
1013		if (rv == -1)
1014			return;
1015
1016		offset += rv;
1017	}
1018
1019done:
1020	m_freem(m);
1021}
1022
1023int
1024pfsync_in_clr(struct pfsync_pkt *pkt, struct mbuf *m, int offset, int count)
1025{
1026	struct pfsync_clr *clr;
1027	struct mbuf *mp;
1028	int len = sizeof(*clr) * count;
1029	int i, offp;
1030
1031	struct pf_state *st, *nexts;
1032	struct pf_state_key *sk, *nextsk;
1033	struct pf_state_item *si;
1034	u_int32_t creatorid;
1035	int s;
1036
1037	mp = m_pulldown(m, offset, len, &offp);
1038	if (mp == NULL) {
1039		V_pfsyncstats.pfsyncs_badlen++;
1040		return (-1);
1041	}
1042	clr = (struct pfsync_clr *)(mp->m_data + offp);
1043
1044	s = splsoftnet();
1045#ifdef __FreeBSD__
1046	PF_LOCK();
1047#endif
1048	for (i = 0; i < count; i++) {
1049		creatorid = clr[i].creatorid;
1050
1051		if (clr[i].ifname[0] == '\0') {
1052#ifdef __FreeBSD__
1053			for (st = RB_MIN(pf_state_tree_id, &V_tree_id);
1054			    st; st = nexts) {
1055				nexts = RB_NEXT(pf_state_tree_id, &V_tree_id, st);
1056#else
1057			for (st = RB_MIN(pf_state_tree_id, &tree_id);
1058			    st; st = nexts) {
1059				nexts = RB_NEXT(pf_state_tree_id, &tree_id, st);
1060#endif
1061				if (st->creatorid == creatorid) {
1062					SET(st->state_flags, PFSTATE_NOSYNC);
1063					pf_unlink_state(st);
1064				}
1065			}
1066		} else {
1067			if (pfi_kif_get(clr[i].ifname) == NULL)
1068				continue;
1069
1070			/* XXX correct? */
1071#ifdef __FreeBSD__
1072			for (sk = RB_MIN(pf_state_tree, &V_pf_statetbl);
1073#else
1074			for (sk = RB_MIN(pf_state_tree, &pf_statetbl);
1075#endif
1076			    sk; sk = nextsk) {
1077				nextsk = RB_NEXT(pf_state_tree,
1078#ifdef __FreeBSD__
1079				    &V_pf_statetbl, sk);
1080#else
1081				    &pf_statetbl, sk);
1082#endif
1083				TAILQ_FOREACH(si, &sk->states, entry) {
1084					if (si->s->creatorid == creatorid) {
1085						SET(si->s->state_flags,
1086						    PFSTATE_NOSYNC);
1087						pf_unlink_state(si->s);
1088					}
1089				}
1090			}
1091		}
1092	}
1093#ifdef __FreeBSD__
1094	PF_UNLOCK();
1095#endif
1096	splx(s);
1097
1098	return (len);
1099}
1100
1101int
1102pfsync_in_ins(struct pfsync_pkt *pkt, struct mbuf *m, int offset, int count)
1103{
1104	struct mbuf *mp;
1105	struct pfsync_state *sa, *sp;
1106	int len = sizeof(*sp) * count;
1107	int i, offp;
1108
1109	int s;
1110
1111	mp = m_pulldown(m, offset, len, &offp);
1112	if (mp == NULL) {
1113		V_pfsyncstats.pfsyncs_badlen++;
1114		return (-1);
1115	}
1116	sa = (struct pfsync_state *)(mp->m_data + offp);
1117
1118	s = splsoftnet();
1119#ifdef __FreeBSD__
1120	PF_LOCK();
1121#endif
1122	for (i = 0; i < count; i++) {
1123		sp = &sa[i];
1124
1125		/* check for invalid values */
1126		if (sp->timeout >= PFTM_MAX ||
1127		    sp->src.state > PF_TCPS_PROXY_DST ||
1128		    sp->dst.state > PF_TCPS_PROXY_DST ||
1129		    sp->direction > PF_OUT ||
1130		    (sp->af != AF_INET && sp->af != AF_INET6)) {
1131#ifdef __FreeBSD__
1132			if (V_pf_status.debug >= PF_DEBUG_MISC) {
1133#else
1134			if (pf_status.debug >= PF_DEBUG_MISC) {
1135#endif
1136				printf("pfsync_input: PFSYNC5_ACT_INS: "
1137				    "invalid value\n");
1138			}
1139			V_pfsyncstats.pfsyncs_badval++;
1140			continue;
1141		}
1142
1143		if (pfsync_state_import(sp, pkt->flags) == ENOMEM) {
1144			/* drop out, but process the rest of the actions */
1145			break;
1146		}
1147	}
1148#ifdef __FreeBSD__
1149	PF_UNLOCK();
1150#endif
1151	splx(s);
1152
1153	return (len);
1154}
1155
1156int
1157pfsync_in_iack(struct pfsync_pkt *pkt, struct mbuf *m, int offset, int count)
1158{
1159	struct pfsync_ins_ack *ia, *iaa;
1160	struct pf_state_cmp id_key;
1161	struct pf_state *st;
1162
1163	struct mbuf *mp;
1164	int len = count * sizeof(*ia);
1165	int offp, i;
1166	int s;
1167
1168	mp = m_pulldown(m, offset, len, &offp);
1169	if (mp == NULL) {
1170		V_pfsyncstats.pfsyncs_badlen++;
1171		return (-1);
1172	}
1173	iaa = (struct pfsync_ins_ack *)(mp->m_data + offp);
1174
1175	s = splsoftnet();
1176#ifdef __FreeBSD__
1177	PF_LOCK();
1178#endif
1179	for (i = 0; i < count; i++) {
1180		ia = &iaa[i];
1181
1182		bcopy(&ia->id, &id_key.id, sizeof(id_key.id));
1183		id_key.creatorid = ia->creatorid;
1184
1185		st = pf_find_state_byid(&id_key);
1186		if (st == NULL)
1187			continue;
1188
1189		if (ISSET(st->state_flags, PFSTATE_ACK))
1190			pfsync_deferred(st, 0);
1191	}
1192#ifdef __FreeBSD__
1193	PF_UNLOCK();
1194#endif
1195	splx(s);
1196	/*
1197	 * XXX this is not yet implemented, but we know the size of the
1198	 * message so we can skip it.
1199	 */
1200
1201	return (count * sizeof(struct pfsync_ins_ack));
1202}
1203
1204int
1205pfsync_upd_tcp(struct pf_state *st, struct pfsync_state_peer *src,
1206    struct pfsync_state_peer *dst)
1207{
1208	int sfail = 0;
1209
1210	/*
1211	 * The state should never go backwards except
1212	 * for syn-proxy states.  Neither should the
1213	 * sequence window slide backwards.
1214	 */
1215	if (st->src.state > src->state &&
1216	    (st->src.state < PF_TCPS_PROXY_SRC ||
1217	    src->state >= PF_TCPS_PROXY_SRC))
1218		sfail = 1;
1219	else if (SEQ_GT(st->src.seqlo, ntohl(src->seqlo)))
1220		sfail = 3;
1221	else if (st->dst.state > dst->state) {
1222		/* There might still be useful
1223		 * information about the src state here,
1224		 * so import that part of the update,
1225		 * then "fail" so we send the updated
1226		 * state back to the peer who is missing
1227		 * our what we know. */
1228		pf_state_peer_ntoh(src, &st->src);
1229		/* XXX do anything with timeouts? */
1230		sfail = 7;
1231	} else if (st->dst.state >= TCPS_SYN_SENT &&
1232	    SEQ_GT(st->dst.seqlo, ntohl(dst->seqlo)))
1233		sfail = 4;
1234
1235	return (sfail);
1236}
1237
1238int
1239pfsync_in_upd(struct pfsync_pkt *pkt, struct mbuf *m, int offset, int count)
1240{
1241	struct pfsync_state *sa, *sp;
1242	struct pf_state_cmp id_key;
1243	struct pf_state_key *sk;
1244	struct pf_state *st;
1245	int sfail;
1246
1247	struct mbuf *mp;
1248	int len = count * sizeof(*sp);
1249	int offp, i;
1250	int s;
1251
1252	mp = m_pulldown(m, offset, len, &offp);
1253	if (mp == NULL) {
1254		V_pfsyncstats.pfsyncs_badlen++;
1255		return (-1);
1256	}
1257	sa = (struct pfsync_state *)(mp->m_data + offp);
1258
1259	s = splsoftnet();
1260#ifdef __FreeBSD__
1261	PF_LOCK();
1262#endif
1263	for (i = 0; i < count; i++) {
1264		sp = &sa[i];
1265
1266		/* check for invalid values */
1267		if (sp->timeout >= PFTM_MAX ||
1268		    sp->src.state > PF_TCPS_PROXY_DST ||
1269		    sp->dst.state > PF_TCPS_PROXY_DST) {
1270#ifdef __FreeBSD__
1271			if (V_pf_status.debug >= PF_DEBUG_MISC) {
1272#else
1273			if (pf_status.debug >= PF_DEBUG_MISC) {
1274#endif
1275				printf("pfsync_input: PFSYNC_ACT_UPD: "
1276				    "invalid value\n");
1277			}
1278			V_pfsyncstats.pfsyncs_badval++;
1279			continue;
1280		}
1281
1282		bcopy(sp->id, &id_key.id, sizeof(id_key.id));
1283		id_key.creatorid = sp->creatorid;
1284
1285		st = pf_find_state_byid(&id_key);
1286		if (st == NULL) {
1287			/* insert the update */
1288			if (pfsync_state_import(sp, 0))
1289				V_pfsyncstats.pfsyncs_badstate++;
1290			continue;
1291		}
1292
1293		if (ISSET(st->state_flags, PFSTATE_ACK))
1294			pfsync_deferred(st, 1);
1295
1296		sk = st->key[PF_SK_WIRE];	/* XXX right one? */
1297		sfail = 0;
1298		if (sk->proto == IPPROTO_TCP)
1299			sfail = pfsync_upd_tcp(st, &sp->src, &sp->dst);
1300		else {
1301			/*
1302			 * Non-TCP protocol state machine always go
1303			 * forwards
1304			 */
1305			if (st->src.state > sp->src.state)
1306				sfail = 5;
1307			else if (st->dst.state > sp->dst.state)
1308				sfail = 6;
1309		}
1310
1311		if (sfail) {
1312#ifdef __FreeBSD__
1313			if (V_pf_status.debug >= PF_DEBUG_MISC) {
1314#else
1315			if (pf_status.debug >= PF_DEBUG_MISC) {
1316#endif
1317				printf("pfsync: %s stale update (%d)"
1318				    " id: %016llx creatorid: %08x\n",
1319				    (sfail < 7 ?  "ignoring" : "partial"),
1320				    sfail, betoh64(st->id),
1321				    ntohl(st->creatorid));
1322			}
1323			V_pfsyncstats.pfsyncs_stale++;
1324
1325			pfsync_update_state(st);
1326#ifdef __FreeBSD__
1327			pfsync_sendout();
1328#else
1329			schednetisr(NETISR_PFSYNC);
1330#endif
1331			continue;
1332		}
1333		pfsync_alloc_scrub_memory(&sp->dst, &st->dst);
1334		pf_state_peer_ntoh(&sp->src, &st->src);
1335		pf_state_peer_ntoh(&sp->dst, &st->dst);
1336		st->expire = ntohl(sp->expire) + time_second;
1337		st->timeout = sp->timeout;
1338		st->pfsync_time = time_uptime;
1339	}
1340#ifdef __FreeBSD__
1341	PF_UNLOCK();
1342#endif
1343	splx(s);
1344
1345	return (len);
1346}
1347
1348int
1349pfsync_in_upd_c(struct pfsync_pkt *pkt, struct mbuf *m, int offset, int count)
1350{
1351	struct pfsync_upd_c *ua, *up;
1352	struct pf_state_key *sk;
1353	struct pf_state_cmp id_key;
1354	struct pf_state *st;
1355
1356	int len = count * sizeof(*up);
1357	int sfail;
1358
1359	struct mbuf *mp;
1360	int offp, i;
1361	int s;
1362
1363	mp = m_pulldown(m, offset, len, &offp);
1364	if (mp == NULL) {
1365		V_pfsyncstats.pfsyncs_badlen++;
1366		return (-1);
1367	}
1368	ua = (struct pfsync_upd_c *)(mp->m_data + offp);
1369
1370	s = splsoftnet();
1371#ifdef __FreeBSD__
1372	PF_LOCK();
1373#endif
1374	for (i = 0; i < count; i++) {
1375		up = &ua[i];
1376
1377		/* check for invalid values */
1378		if (up->timeout >= PFTM_MAX ||
1379		    up->src.state > PF_TCPS_PROXY_DST ||
1380		    up->dst.state > PF_TCPS_PROXY_DST) {
1381#ifdef __FreeBSD__
1382			if (V_pf_status.debug >= PF_DEBUG_MISC) {
1383#else
1384			if (pf_status.debug >= PF_DEBUG_MISC) {
1385#endif
1386				printf("pfsync_input: "
1387				    "PFSYNC_ACT_UPD_C: "
1388				    "invalid value\n");
1389			}
1390			V_pfsyncstats.pfsyncs_badval++;
1391			continue;
1392		}
1393
1394		bcopy(&up->id, &id_key.id, sizeof(id_key.id));
1395		id_key.creatorid = up->creatorid;
1396
1397		st = pf_find_state_byid(&id_key);
1398		if (st == NULL) {
1399			/* We don't have this state. Ask for it. */
1400			pfsync_request_update(id_key.creatorid, id_key.id);
1401			continue;
1402		}
1403
1404		if (ISSET(st->state_flags, PFSTATE_ACK))
1405			pfsync_deferred(st, 1);
1406
1407		sk = st->key[PF_SK_WIRE]; /* XXX right one? */
1408		sfail = 0;
1409		if (sk->proto == IPPROTO_TCP)
1410			sfail = pfsync_upd_tcp(st, &up->src, &up->dst);
1411		else {
1412			/*
1413			 * Non-TCP protocol state machine always go forwards
1414			 */
1415			if (st->src.state > up->src.state)
1416				sfail = 5;
1417			else if (st->dst.state > up->dst.state)
1418				sfail = 6;
1419		}
1420
1421		if (sfail) {
1422#ifdef __FreeBSD__
1423			if (V_pf_status.debug >= PF_DEBUG_MISC) {
1424#else
1425			if (pf_status.debug >= PF_DEBUG_MISC) {
1426#endif
1427				printf("pfsync: ignoring stale update "
1428				    "(%d) id: %016llx "
1429				    "creatorid: %08x\n", sfail,
1430				    betoh64(st->id),
1431				    ntohl(st->creatorid));
1432			}
1433			V_pfsyncstats.pfsyncs_stale++;
1434
1435			pfsync_update_state(st);
1436#ifdef __FreeBSD__
1437			pfsync_sendout();
1438#else
1439			schednetisr(NETISR_PFSYNC);
1440#endif
1441			continue;
1442		}
1443		pfsync_alloc_scrub_memory(&up->dst, &st->dst);
1444		pf_state_peer_ntoh(&up->src, &st->src);
1445		pf_state_peer_ntoh(&up->dst, &st->dst);
1446		st->expire = ntohl(up->expire) + time_second;
1447		st->timeout = up->timeout;
1448		st->pfsync_time = time_uptime;
1449	}
1450#ifdef __FreeBSD__
1451	PF_UNLOCK();
1452#endif
1453	splx(s);
1454
1455	return (len);
1456}
1457
1458int
1459pfsync_in_ureq(struct pfsync_pkt *pkt, struct mbuf *m, int offset, int count)
1460{
1461	struct pfsync_upd_req *ur, *ura;
1462	struct mbuf *mp;
1463	int len = count * sizeof(*ur);
1464	int i, offp;
1465
1466	struct pf_state_cmp id_key;
1467	struct pf_state *st;
1468
1469	mp = m_pulldown(m, offset, len, &offp);
1470	if (mp == NULL) {
1471		V_pfsyncstats.pfsyncs_badlen++;
1472		return (-1);
1473	}
1474	ura = (struct pfsync_upd_req *)(mp->m_data + offp);
1475
1476	for (i = 0; i < count; i++) {
1477		ur = &ura[i];
1478
1479		bcopy(&ur->id, &id_key.id, sizeof(id_key.id));
1480		id_key.creatorid = ur->creatorid;
1481
1482		if (id_key.id == 0 && id_key.creatorid == 0)
1483			pfsync_bulk_start();
1484		else {
1485			st = pf_find_state_byid(&id_key);
1486			if (st == NULL) {
1487				V_pfsyncstats.pfsyncs_badstate++;
1488				continue;
1489			}
1490			if (ISSET(st->state_flags, PFSTATE_NOSYNC))
1491				continue;
1492
1493			PF_LOCK();
1494			pfsync_update_state_req(st);
1495			PF_UNLOCK();
1496		}
1497	}
1498
1499	return (len);
1500}
1501
1502int
1503pfsync_in_del(struct pfsync_pkt *pkt, struct mbuf *m, int offset, int count)
1504{
1505	struct mbuf *mp;
1506	struct pfsync_state *sa, *sp;
1507	struct pf_state_cmp id_key;
1508	struct pf_state *st;
1509	int len = count * sizeof(*sp);
1510	int offp, i;
1511	int s;
1512
1513	mp = m_pulldown(m, offset, len, &offp);
1514	if (mp == NULL) {
1515		V_pfsyncstats.pfsyncs_badlen++;
1516		return (-1);
1517	}
1518	sa = (struct pfsync_state *)(mp->m_data + offp);
1519
1520	s = splsoftnet();
1521#ifdef __FreeBSD__
1522	PF_LOCK();
1523#endif
1524	for (i = 0; i < count; i++) {
1525		sp = &sa[i];
1526
1527		bcopy(sp->id, &id_key.id, sizeof(id_key.id));
1528		id_key.creatorid = sp->creatorid;
1529
1530		st = pf_find_state_byid(&id_key);
1531		if (st == NULL) {
1532			V_pfsyncstats.pfsyncs_badstate++;
1533			continue;
1534		}
1535		SET(st->state_flags, PFSTATE_NOSYNC);
1536		pf_unlink_state(st);
1537	}
1538#ifdef __FreeBSD__
1539	PF_UNLOCK();
1540#endif
1541	splx(s);
1542
1543	return (len);
1544}
1545
1546int
1547pfsync_in_del_c(struct pfsync_pkt *pkt, struct mbuf *m, int offset, int count)
1548{
1549	struct mbuf *mp;
1550	struct pfsync_del_c *sa, *sp;
1551	struct pf_state_cmp id_key;
1552	struct pf_state *st;
1553	int len = count * sizeof(*sp);
1554	int offp, i;
1555	int s;
1556
1557	mp = m_pulldown(m, offset, len, &offp);
1558	if (mp == NULL) {
1559		V_pfsyncstats.pfsyncs_badlen++;
1560		return (-1);
1561	}
1562	sa = (struct pfsync_del_c *)(mp->m_data + offp);
1563
1564	s = splsoftnet();
1565#ifdef __FreeBSD__
1566	PF_LOCK();
1567#endif
1568	for (i = 0; i < count; i++) {
1569		sp = &sa[i];
1570
1571		bcopy(&sp->id, &id_key.id, sizeof(id_key.id));
1572		id_key.creatorid = sp->creatorid;
1573
1574		st = pf_find_state_byid(&id_key);
1575		if (st == NULL) {
1576			V_pfsyncstats.pfsyncs_badstate++;
1577			continue;
1578		}
1579
1580		SET(st->state_flags, PFSTATE_NOSYNC);
1581		pf_unlink_state(st);
1582	}
1583#ifdef __FreeBSD__
1584	PF_UNLOCK();
1585#endif
1586	splx(s);
1587
1588	return (len);
1589}
1590
1591int
1592pfsync_in_bus(struct pfsync_pkt *pkt, struct mbuf *m, int offset, int count)
1593{
1594#ifdef __FreeBSD__
1595	struct pfsync_softc *sc = V_pfsyncif;
1596#else
1597	struct pfsync_softc *sc = pfsyncif;
1598#endif
1599	struct pfsync_bus *bus;
1600	struct mbuf *mp;
1601	int len = count * sizeof(*bus);
1602	int offp;
1603
1604	/* If we're not waiting for a bulk update, who cares. */
1605	if (sc->sc_ureq_sent == 0)
1606		return (len);
1607
1608	mp = m_pulldown(m, offset, len, &offp);
1609	if (mp == NULL) {
1610		V_pfsyncstats.pfsyncs_badlen++;
1611		return (-1);
1612	}
1613	bus = (struct pfsync_bus *)(mp->m_data + offp);
1614
1615	switch (bus->status) {
1616	case PFSYNC_BUS_START:
1617#ifdef __FreeBSD__
1618		callout_reset(&sc->sc_bulkfail_tmo, 4 * hz +
1619		    V_pf_pool_limits[PF_LIMIT_STATES].limit /
1620		    ((sc->sc_sync_if->if_mtu - PFSYNC_MINPKT) /
1621		    sizeof(struct pfsync_state)),
1622		    pfsync_bulk_fail, V_pfsyncif);
1623#else
1624		timeout_add(&sc->sc_bulkfail_tmo, 4 * hz +
1625		    pf_pool_limits[PF_LIMIT_STATES].limit /
1626		    ((sc->sc_if.if_mtu - PFSYNC_MINPKT) /
1627		    sizeof(struct pfsync_state)));
1628#endif
1629#ifdef __FreeBSD__
1630		if (V_pf_status.debug >= PF_DEBUG_MISC)
1631#else
1632		if (pf_status.debug >= PF_DEBUG_MISC)
1633#endif
1634			printf("pfsync: received bulk update start\n");
1635		break;
1636
1637	case PFSYNC_BUS_END:
1638		if (time_uptime - ntohl(bus->endtime) >=
1639		    sc->sc_ureq_sent) {
1640			/* that's it, we're happy */
1641			sc->sc_ureq_sent = 0;
1642			sc->sc_bulk_tries = 0;
1643			timeout_del(&sc->sc_bulkfail_tmo);
1644#if NCARP > 0
1645#ifdef notyet
1646#ifdef __FreeBSD__
1647			if (!sc->pfsync_sync_ok)
1648#else
1649			if (!pfsync_sync_ok)
1650#endif
1651				carp_group_demote_adj(&sc->sc_if, -1);
1652#endif
1653#endif
1654#ifdef __FreeBSD__
1655			sc->pfsync_sync_ok = 1;
1656#else
1657			pfsync_sync_ok = 1;
1658#endif
1659#ifdef __FreeBSD__
1660			if (V_pf_status.debug >= PF_DEBUG_MISC)
1661#else
1662			if (pf_status.debug >= PF_DEBUG_MISC)
1663#endif
1664				printf("pfsync: received valid "
1665				    "bulk update end\n");
1666		} else {
1667#ifdef __FreeBSD__
1668			if (V_pf_status.debug >= PF_DEBUG_MISC)
1669#else
1670			if (pf_status.debug >= PF_DEBUG_MISC)
1671#endif
1672				printf("pfsync: received invalid "
1673				    "bulk update end: bad timestamp\n");
1674		}
1675		break;
1676	}
1677
1678	return (len);
1679}
1680
1681int
1682pfsync_in_tdb(struct pfsync_pkt *pkt, struct mbuf *m, int offset, int count)
1683{
1684	int len = count * sizeof(struct pfsync_tdb);
1685
1686#if defined(IPSEC)
1687	struct pfsync_tdb *tp;
1688	struct mbuf *mp;
1689	int offp;
1690	int i;
1691	int s;
1692
1693	mp = m_pulldown(m, offset, len, &offp);
1694	if (mp == NULL) {
1695		V_pfsyncstats.pfsyncs_badlen++;
1696		return (-1);
1697	}
1698	tp = (struct pfsync_tdb *)(mp->m_data + offp);
1699
1700	s = splsoftnet();
1701#ifdef __FreeBSD__
1702	PF_LOCK();
1703#endif
1704	for (i = 0; i < count; i++)
1705		pfsync_update_net_tdb(&tp[i]);
1706#ifdef __FreeBSD__
1707	PF_UNLOCK();
1708#endif
1709	splx(s);
1710#endif
1711
1712	return (len);
1713}
1714
1715#if defined(IPSEC)
1716/* Update an in-kernel tdb. Silently fail if no tdb is found. */
1717void
1718pfsync_update_net_tdb(struct pfsync_tdb *pt)
1719{
1720	struct tdb		*tdb;
1721	int			 s;
1722
1723	/* check for invalid values */
1724	if (ntohl(pt->spi) <= SPI_RESERVED_MAX ||
1725	    (pt->dst.sa.sa_family != AF_INET &&
1726	     pt->dst.sa.sa_family != AF_INET6))
1727		goto bad;
1728
1729	s = spltdb();
1730	tdb = gettdb(pt->spi, &pt->dst, pt->sproto);
1731	if (tdb) {
1732		pt->rpl = ntohl(pt->rpl);
1733		pt->cur_bytes = betoh64(pt->cur_bytes);
1734
1735		/* Neither replay nor byte counter should ever decrease. */
1736		if (pt->rpl < tdb->tdb_rpl ||
1737		    pt->cur_bytes < tdb->tdb_cur_bytes) {
1738			splx(s);
1739			goto bad;
1740		}
1741
1742		tdb->tdb_rpl = pt->rpl;
1743		tdb->tdb_cur_bytes = pt->cur_bytes;
1744	}
1745	splx(s);
1746	return;
1747
1748bad:
1749#ifdef __FreeBSD__
1750	if (V_pf_status.debug >= PF_DEBUG_MISC)
1751#else
1752	if (pf_status.debug >= PF_DEBUG_MISC)
1753#endif
1754		printf("pfsync_insert: PFSYNC_ACT_TDB_UPD: "
1755		    "invalid value\n");
1756	V_pfsyncstats.pfsyncs_badstate++;
1757	return;
1758}
1759#endif
1760
1761
1762int
1763pfsync_in_eof(struct pfsync_pkt *pkt, struct mbuf *m, int offset, int count)
1764{
1765	/* check if we are at the right place in the packet */
1766	if (offset != m->m_pkthdr.len - sizeof(struct pfsync_eof))
1767		V_pfsyncstats.pfsyncs_badact++;
1768
1769	/* we're done. free and let the caller return */
1770	m_freem(m);
1771	return (-1);
1772}
1773
1774int
1775pfsync_in_error(struct pfsync_pkt *pkt, struct mbuf *m, int offset, int count)
1776{
1777	V_pfsyncstats.pfsyncs_badact++;
1778
1779	m_freem(m);
1780	return (-1);
1781}
1782
1783int
1784pfsyncoutput(struct ifnet *ifp, struct mbuf *m, struct sockaddr *dst,
1785#ifdef __FreeBSD__
1786	struct route *rt)
1787#else
1788	struct rtentry *rt)
1789#endif
1790{
1791	m_freem(m);
1792	return (0);
1793}
1794
1795/* ARGSUSED */
1796int
1797pfsyncioctl(struct ifnet *ifp, u_long cmd, caddr_t data)
1798{
1799#ifndef __FreeBSD__
1800	struct proc *p = curproc;
1801#endif
1802	struct pfsync_softc *sc = ifp->if_softc;
1803	struct ifreq *ifr = (struct ifreq *)data;
1804	struct ip_moptions *imo = &sc->sc_imo;
1805	struct pfsyncreq pfsyncr;
1806	struct ifnet    *sifp;
1807	struct ip *ip;
1808	int s, error;
1809
1810	switch (cmd) {
1811#if 0
1812	case SIOCSIFADDR:
1813	case SIOCAIFADDR:
1814	case SIOCSIFDSTADDR:
1815#endif
1816	case SIOCSIFFLAGS:
1817#ifdef __FreeBSD__
1818		if (ifp->if_flags & IFF_UP)
1819			ifp->if_drv_flags |= IFF_DRV_RUNNING;
1820		else
1821			ifp->if_drv_flags &= ~IFF_DRV_RUNNING;
1822#else
1823		if (ifp->if_flags & IFF_UP)
1824			ifp->if_flags |= IFF_RUNNING;
1825		else
1826			ifp->if_flags &= ~IFF_RUNNING;
1827#endif
1828		break;
1829	case SIOCSIFMTU:
1830		if (ifr->ifr_mtu <= PFSYNC_MINPKT)
1831			return (EINVAL);
1832		if (ifr->ifr_mtu > MCLBYTES) /* XXX could be bigger */
1833			ifr->ifr_mtu = MCLBYTES;
1834		if (ifr->ifr_mtu < ifp->if_mtu) {
1835			s = splnet();
1836#ifdef __FreeBSD__
1837			PF_LOCK();
1838#endif
1839			pfsync_sendout();
1840#ifdef __FreeBSD__
1841			PF_UNLOCK();
1842#endif
1843			splx(s);
1844		}
1845		ifp->if_mtu = ifr->ifr_mtu;
1846		break;
1847	case SIOCGETPFSYNC:
1848		bzero(&pfsyncr, sizeof(pfsyncr));
1849		if (sc->sc_sync_if) {
1850			strlcpy(pfsyncr.pfsyncr_syncdev,
1851			    sc->sc_sync_if->if_xname, IFNAMSIZ);
1852		}
1853		pfsyncr.pfsyncr_syncpeer = sc->sc_sync_peer;
1854		pfsyncr.pfsyncr_maxupdates = sc->sc_maxupdates;
1855		return (copyout(&pfsyncr, ifr->ifr_data, sizeof(pfsyncr)));
1856
1857	case SIOCSETPFSYNC:
1858#ifdef __FreeBSD__
1859		if ((error = priv_check(curthread, PRIV_NETINET_PF)) != 0)
1860#else
1861		if ((error = suser(p, p->p_acflag)) != 0)
1862#endif
1863			return (error);
1864		if ((error = copyin(ifr->ifr_data, &pfsyncr, sizeof(pfsyncr))))
1865			return (error);
1866
1867#ifdef __FreeBSD__
1868		PF_LOCK();
1869#endif
1870		if (pfsyncr.pfsyncr_syncpeer.s_addr == 0)
1871#ifdef __FreeBSD__
1872			sc->sc_sync_peer.s_addr = htonl(INADDR_PFSYNC_GROUP);
1873#else
1874			sc->sc_sync_peer.s_addr = INADDR_PFSYNC_GROUP;
1875#endif
1876		else
1877			sc->sc_sync_peer.s_addr =
1878			    pfsyncr.pfsyncr_syncpeer.s_addr;
1879
1880		if (pfsyncr.pfsyncr_maxupdates > 255)
1881#ifdef __FreeBSD__
1882		{
1883			PF_UNLOCK();
1884#endif
1885			return (EINVAL);
1886#ifdef __FreeBSD__
1887		}
1888#endif
1889		sc->sc_maxupdates = pfsyncr.pfsyncr_maxupdates;
1890
1891		if (pfsyncr.pfsyncr_syncdev[0] == 0) {
1892			sc->sc_sync_if = NULL;
1893#ifdef __FreeBSD__
1894			PF_UNLOCK();
1895#endif
1896			if (imo->imo_num_memberships > 0) {
1897				in_delmulti(imo->imo_membership[
1898				    --imo->imo_num_memberships]);
1899				imo->imo_multicast_ifp = NULL;
1900			}
1901			break;
1902		}
1903
1904#ifdef __FreeBSD__
1905		PF_UNLOCK();
1906#endif
1907		if ((sifp = ifunit(pfsyncr.pfsyncr_syncdev)) == NULL)
1908			return (EINVAL);
1909
1910#ifdef __FreeBSD__
1911		PF_LOCK();
1912#endif
1913		s = splnet();
1914#ifdef __FreeBSD__
1915		if (sifp->if_mtu < sc->sc_ifp->if_mtu ||
1916#else
1917		if (sifp->if_mtu < sc->sc_if.if_mtu ||
1918#endif
1919		    (sc->sc_sync_if != NULL &&
1920		    sifp->if_mtu < sc->sc_sync_if->if_mtu) ||
1921		    sifp->if_mtu < MCLBYTES - sizeof(struct ip))
1922			pfsync_sendout();
1923		sc->sc_sync_if = sifp;
1924
1925		if (imo->imo_num_memberships > 0) {
1926#ifdef __FreeBSD__
1927			PF_UNLOCK();
1928#endif
1929			in_delmulti(imo->imo_membership[--imo->imo_num_memberships]);
1930#ifdef __FreeBSD__
1931			PF_LOCK();
1932#endif
1933			imo->imo_multicast_ifp = NULL;
1934		}
1935
1936		if (sc->sc_sync_if &&
1937#ifdef __FreeBSD__
1938		    sc->sc_sync_peer.s_addr == htonl(INADDR_PFSYNC_GROUP)) {
1939#else
1940		    sc->sc_sync_peer.s_addr == INADDR_PFSYNC_GROUP) {
1941#endif
1942			struct in_addr addr;
1943
1944			if (!(sc->sc_sync_if->if_flags & IFF_MULTICAST)) {
1945				sc->sc_sync_if = NULL;
1946#ifdef __FreeBSD__
1947				PF_UNLOCK();
1948#endif
1949				splx(s);
1950				return (EADDRNOTAVAIL);
1951			}
1952
1953#ifdef __FreeBSD__
1954			addr.s_addr = htonl(INADDR_PFSYNC_GROUP);
1955#else
1956			addr.s_addr = INADDR_PFSYNC_GROUP;
1957#endif
1958
1959#ifdef __FreeBSD__
1960			PF_UNLOCK();
1961#endif
1962			if ((imo->imo_membership[0] =
1963			    in_addmulti(&addr, sc->sc_sync_if)) == NULL) {
1964				sc->sc_sync_if = NULL;
1965				splx(s);
1966				return (ENOBUFS);
1967			}
1968#ifdef __FreeBSD__
1969			PF_LOCK();
1970#endif
1971			imo->imo_num_memberships++;
1972			imo->imo_multicast_ifp = sc->sc_sync_if;
1973			imo->imo_multicast_ttl = PFSYNC_DFLTTL;
1974			imo->imo_multicast_loop = 0;
1975		}
1976
1977		ip = &sc->sc_template;
1978		bzero(ip, sizeof(*ip));
1979		ip->ip_v = IPVERSION;
1980		ip->ip_hl = sizeof(sc->sc_template) >> 2;
1981		ip->ip_tos = IPTOS_LOWDELAY;
1982		/* len and id are set later */
1983#ifdef __FreeBSD__
1984		ip->ip_off = IP_DF;
1985#else
1986		ip->ip_off = htons(IP_DF);
1987#endif
1988		ip->ip_ttl = PFSYNC_DFLTTL;
1989		ip->ip_p = IPPROTO_PFSYNC;
1990		ip->ip_src.s_addr = INADDR_ANY;
1991		ip->ip_dst.s_addr = sc->sc_sync_peer.s_addr;
1992
1993		if (sc->sc_sync_if) {
1994			/* Request a full state table update. */
1995			sc->sc_ureq_sent = time_uptime;
1996#if NCARP > 0
1997#ifdef notyet
1998#ifdef __FreeBSD__
1999			if (sc->pfsync_sync_ok)
2000#else
2001			if (pfsync_sync_ok)
2002#endif
2003				carp_group_demote_adj(&sc->sc_if, 1);
2004#endif
2005#endif
2006#ifdef __FreeBSD__
2007			sc->pfsync_sync_ok = 0;
2008#else
2009			pfsync_sync_ok = 0;
2010#endif
2011#ifdef __FreeBSD__
2012			if (V_pf_status.debug >= PF_DEBUG_MISC)
2013#else
2014			if (pf_status.debug >= PF_DEBUG_MISC)
2015#endif
2016				printf("pfsync: requesting bulk update\n");
2017#ifdef __FreeBSD__
2018			callout_reset(&sc->sc_bulkfail_tmo, 5 * hz,
2019			    pfsync_bulk_fail, V_pfsyncif);
2020#else
2021			timeout_add_sec(&sc->sc_bulkfail_tmo, 5);
2022#endif
2023			pfsync_request_update(0, 0);
2024		}
2025#ifdef __FreeBSD__
2026		PF_UNLOCK();
2027#endif
2028		splx(s);
2029
2030		break;
2031
2032	default:
2033		return (ENOTTY);
2034	}
2035
2036	return (0);
2037}
2038
2039int
2040pfsync_out_state(struct pf_state *st, struct mbuf *m, int offset)
2041{
2042	struct pfsync_state *sp = (struct pfsync_state *)(m->m_data + offset);
2043
2044	pfsync_state_export(sp, st);
2045
2046	return (sizeof(*sp));
2047}
2048
2049int
2050pfsync_out_iack(struct pf_state *st, struct mbuf *m, int offset)
2051{
2052	struct pfsync_ins_ack *iack =
2053	    (struct pfsync_ins_ack *)(m->m_data + offset);
2054
2055	iack->id = st->id;
2056	iack->creatorid = st->creatorid;
2057
2058	return (sizeof(*iack));
2059}
2060
2061int
2062pfsync_out_upd_c(struct pf_state *st, struct mbuf *m, int offset)
2063{
2064	struct pfsync_upd_c *up = (struct pfsync_upd_c *)(m->m_data + offset);
2065
2066	up->id = st->id;
2067	pf_state_peer_hton(&st->src, &up->src);
2068	pf_state_peer_hton(&st->dst, &up->dst);
2069	up->creatorid = st->creatorid;
2070
2071	up->expire = pf_state_expires(st);
2072	if (up->expire <= time_second)
2073		up->expire = htonl(0);
2074	else
2075		up->expire = htonl(up->expire - time_second);
2076	up->timeout = st->timeout;
2077
2078	bzero(up->_pad, sizeof(up->_pad)); /* XXX */
2079
2080	return (sizeof(*up));
2081}
2082
2083int
2084pfsync_out_del(struct pf_state *st, struct mbuf *m, int offset)
2085{
2086	struct pfsync_del_c *dp = (struct pfsync_del_c *)(m->m_data + offset);
2087
2088	dp->id = st->id;
2089	dp->creatorid = st->creatorid;
2090
2091	SET(st->state_flags, PFSTATE_NOSYNC);
2092
2093	return (sizeof(*dp));
2094}
2095
2096void
2097pfsync_drop(struct pfsync_softc *sc)
2098{
2099	struct pf_state *st;
2100	struct pfsync_upd_req_item *ur;
2101#ifdef notyet
2102	struct tdb *t;
2103#endif
2104	int q;
2105
2106	for (q = 0; q < PFSYNC_S_COUNT; q++) {
2107		if (TAILQ_EMPTY(&sc->sc_qs[q]))
2108			continue;
2109
2110		TAILQ_FOREACH(st, &sc->sc_qs[q], sync_list) {
2111#ifdef PFSYNC_DEBUG
2112#ifdef __FreeBSD__
2113			KASSERT(st->sync_state == q,
2114				("%s: st->sync_state == q",
2115					__FUNCTION__));
2116#else
2117			KASSERT(st->sync_state == q);
2118#endif
2119#endif
2120			st->sync_state = PFSYNC_S_NONE;
2121		}
2122		TAILQ_INIT(&sc->sc_qs[q]);
2123	}
2124
2125	while ((ur = TAILQ_FIRST(&sc->sc_upd_req_list)) != NULL) {
2126		TAILQ_REMOVE(&sc->sc_upd_req_list, ur, ur_entry);
2127		pool_put(&sc->sc_pool, ur);
2128	}
2129
2130	sc->sc_plus = NULL;
2131
2132#ifdef notyet
2133	if (!TAILQ_EMPTY(&sc->sc_tdb_q)) {
2134		TAILQ_FOREACH(t, &sc->sc_tdb_q, tdb_sync_entry)
2135			CLR(t->tdb_flags, TDBF_PFSYNC);
2136
2137		TAILQ_INIT(&sc->sc_tdb_q);
2138	}
2139#endif
2140
2141	sc->sc_len = PFSYNC_MINPKT;
2142}
2143
2144void
2145pfsync_sendout(void)
2146{
2147#ifdef __FreeBSD__
2148	struct pfsync_softc *sc = V_pfsyncif;
2149#else
2150	struct pfsync_softc *sc = pfsyncif;
2151#endif
2152#if NBPFILTER > 0
2153#ifdef __FreeBSD__
2154	struct ifnet *ifp = sc->sc_ifp;
2155#else
2156	struct ifnet *ifp = &sc->sc_if;
2157#endif
2158#endif
2159	struct mbuf *m;
2160	struct ip *ip;
2161	struct pfsync_header *ph;
2162	struct pfsync_subheader *subh;
2163	struct pf_state *st;
2164	struct pfsync_upd_req_item *ur;
2165#ifdef notyet
2166	struct tdb *t;
2167#endif
2168#ifdef __FreeBSD__
2169	size_t pktlen;
2170	int dummy_error;
2171#endif
2172	int offset;
2173	int q, count = 0;
2174
2175#ifdef __FreeBSD__
2176	PF_LOCK_ASSERT();
2177#else
2178	splassert(IPL_NET);
2179#endif
2180
2181	if (sc == NULL || sc->sc_len == PFSYNC_MINPKT)
2182		return;
2183
2184#if NBPFILTER > 0
2185	if (ifp->if_bpf == NULL && sc->sc_sync_if == NULL) {
2186#else
2187	if (sc->sc_sync_if == NULL) {
2188#endif
2189		pfsync_drop(sc);
2190		return;
2191	}
2192
2193	MGETHDR(m, M_DONTWAIT, MT_DATA);
2194	if (m == NULL) {
2195#ifdef __FreeBSD__
2196		sc->sc_ifp->if_oerrors++;
2197#else
2198		sc->sc_if.if_oerrors++;
2199#endif
2200		V_pfsyncstats.pfsyncs_onomem++;
2201		pfsync_drop(sc);
2202		return;
2203	}
2204
2205#ifdef __FreeBSD__
2206	pktlen = max_linkhdr + sc->sc_len;
2207	if (pktlen > MHLEN) {
2208		/* Find the right pool to allocate from. */
2209		/* XXX: This is ugly. */
2210		m_cljget(m, M_DONTWAIT, pktlen <= MSIZE ? MSIZE :
2211			pktlen <= MCLBYTES ? MCLBYTES :
2212#if MJUMPAGESIZE != MCLBYTES
2213			pktlen <= MJUMPAGESIZE ? MJUMPAGESIZE :
2214#endif
2215			pktlen <= MJUM9BYTES ? MJUM9BYTES : MJUM16BYTES);
2216#else
2217	if (max_linkhdr + sc->sc_len > MHLEN) {
2218		MCLGETI(m, M_DONTWAIT, NULL, max_linkhdr + sc->sc_len);
2219#endif
2220		if (!ISSET(m->m_flags, M_EXT)) {
2221			m_free(m);
2222#ifdef __FreeBSD__
2223			sc->sc_ifp->if_oerrors++;
2224#else
2225			sc->sc_if.if_oerrors++;
2226#endif
2227			V_pfsyncstats.pfsyncs_onomem++;
2228			pfsync_drop(sc);
2229			return;
2230		}
2231	}
2232	m->m_data += max_linkhdr;
2233	m->m_len = m->m_pkthdr.len = sc->sc_len;
2234
2235	/* build the ip header */
2236	ip = (struct ip *)m->m_data;
2237	bcopy(&sc->sc_template, ip, sizeof(*ip));
2238	offset = sizeof(*ip);
2239
2240#ifdef __FreeBSD__
2241	ip->ip_len = m->m_pkthdr.len;
2242#else
2243	ip->ip_len = htons(m->m_pkthdr.len);
2244#endif
2245	ip->ip_id = htons(ip_randomid());
2246
2247	/* build the pfsync header */
2248	ph = (struct pfsync_header *)(m->m_data + offset);
2249	bzero(ph, sizeof(*ph));
2250	offset += sizeof(*ph);
2251
2252	ph->version = PFSYNC_VERSION;
2253	ph->len = htons(sc->sc_len - sizeof(*ip));
2254#ifdef __FreeBSD__
2255	bcopy(V_pf_status.pf_chksum, ph->pfcksum, PF_MD5_DIGEST_LENGTH);
2256#else
2257	bcopy(pf_status.pf_chksum, ph->pfcksum, PF_MD5_DIGEST_LENGTH);
2258#endif
2259
2260	/* walk the queues */
2261	for (q = 0; q < PFSYNC_S_COUNT; q++) {
2262		if (TAILQ_EMPTY(&sc->sc_qs[q]))
2263			continue;
2264
2265		subh = (struct pfsync_subheader *)(m->m_data + offset);
2266		offset += sizeof(*subh);
2267
2268		count = 0;
2269		TAILQ_FOREACH(st, &sc->sc_qs[q], sync_list) {
2270#ifdef PFSYNC_DEBUG
2271#ifdef __FreeBSD__
2272			KASSERT(st->sync_state == q,
2273				("%s: st->sync_state == q",
2274					__FUNCTION__));
2275#else
2276			KASSERT(st->sync_state == q);
2277#endif
2278#endif
2279
2280			offset += pfsync_qs[q].write(st, m, offset);
2281			st->sync_state = PFSYNC_S_NONE;
2282			count++;
2283		}
2284		TAILQ_INIT(&sc->sc_qs[q]);
2285
2286		bzero(subh, sizeof(*subh));
2287		subh->action = pfsync_qs[q].action;
2288		subh->count = htons(count);
2289	}
2290
2291	if (!TAILQ_EMPTY(&sc->sc_upd_req_list)) {
2292		subh = (struct pfsync_subheader *)(m->m_data + offset);
2293		offset += sizeof(*subh);
2294
2295		count = 0;
2296		while ((ur = TAILQ_FIRST(&sc->sc_upd_req_list)) != NULL) {
2297			TAILQ_REMOVE(&sc->sc_upd_req_list, ur, ur_entry);
2298
2299			bcopy(&ur->ur_msg, m->m_data + offset,
2300			    sizeof(ur->ur_msg));
2301			offset += sizeof(ur->ur_msg);
2302
2303			pool_put(&sc->sc_pool, ur);
2304
2305			count++;
2306		}
2307
2308		bzero(subh, sizeof(*subh));
2309		subh->action = PFSYNC_ACT_UPD_REQ;
2310		subh->count = htons(count);
2311	}
2312
2313	/* has someone built a custom region for us to add? */
2314	if (sc->sc_plus != NULL) {
2315		bcopy(sc->sc_plus, m->m_data + offset, sc->sc_pluslen);
2316		offset += sc->sc_pluslen;
2317
2318		sc->sc_plus = NULL;
2319	}
2320
2321#ifdef notyet
2322	if (!TAILQ_EMPTY(&sc->sc_tdb_q)) {
2323		subh = (struct pfsync_subheader *)(m->m_data + offset);
2324		offset += sizeof(*subh);
2325
2326		count = 0;
2327		TAILQ_FOREACH(t, &sc->sc_tdb_q, tdb_sync_entry) {
2328			offset += pfsync_out_tdb(t, m, offset);
2329			CLR(t->tdb_flags, TDBF_PFSYNC);
2330
2331			count++;
2332		}
2333		TAILQ_INIT(&sc->sc_tdb_q);
2334
2335		bzero(subh, sizeof(*subh));
2336		subh->action = PFSYNC_ACT_TDB;
2337		subh->count = htons(count);
2338	}
2339#endif
2340
2341	subh = (struct pfsync_subheader *)(m->m_data + offset);
2342	offset += sizeof(*subh);
2343
2344	bzero(subh, sizeof(*subh));
2345	subh->action = PFSYNC_ACT_EOF;
2346	subh->count = htons(1);
2347
2348	/* XXX write checksum in EOF here */
2349
2350	/* we're done, let's put it on the wire */
2351#if NBPFILTER > 0
2352	if (ifp->if_bpf) {
2353		m->m_data += sizeof(*ip);
2354		m->m_len = m->m_pkthdr.len = sc->sc_len - sizeof(*ip);
2355#ifdef __FreeBSD__
2356		BPF_MTAP(ifp, m);
2357#else
2358		bpf_mtap(ifp->if_bpf, m, BPF_DIRECTION_OUT);
2359#endif
2360		m->m_data -= sizeof(*ip);
2361		m->m_len = m->m_pkthdr.len = sc->sc_len;
2362	}
2363
2364	if (sc->sc_sync_if == NULL) {
2365		sc->sc_len = PFSYNC_MINPKT;
2366		m_freem(m);
2367		return;
2368	}
2369#endif
2370
2371#ifdef __FreeBSD__
2372	sc->sc_ifp->if_opackets++;
2373	sc->sc_ifp->if_obytes += m->m_pkthdr.len;
2374	sc->sc_len = PFSYNC_MINPKT;
2375
2376	IFQ_ENQUEUE(&sc->sc_ifp->if_snd, m, dummy_error);
2377	schednetisr(NETISR_PFSYNC);
2378#else
2379	sc->sc_if.if_opackets++;
2380	sc->sc_if.if_obytes += m->m_pkthdr.len;
2381
2382	if (ip_output(m, NULL, NULL, IP_RAWOUTPUT, &sc->sc_imo, NULL) == 0)
2383		pfsyncstats.pfsyncs_opackets++;
2384	else
2385		pfsyncstats.pfsyncs_oerrors++;
2386
2387	/* start again */
2388	sc->sc_len = PFSYNC_MINPKT;
2389#endif
2390}
2391
2392void
2393pfsync_insert_state(struct pf_state *st)
2394{
2395#ifdef __FreeBSD__
2396	struct pfsync_softc *sc = V_pfsyncif;
2397#else
2398	struct pfsync_softc *sc = pfsyncif;
2399#endif
2400
2401#ifdef __FreeBSD__
2402	PF_LOCK_ASSERT();
2403#else
2404	splassert(IPL_SOFTNET);
2405#endif
2406
2407	if (ISSET(st->rule.ptr->rule_flag, PFRULE_NOSYNC) ||
2408	    st->key[PF_SK_WIRE]->proto == IPPROTO_PFSYNC) {
2409		SET(st->state_flags, PFSTATE_NOSYNC);
2410		return;
2411	}
2412
2413	if (sc == NULL || ISSET(st->state_flags, PFSTATE_NOSYNC))
2414		return;
2415
2416#ifdef PFSYNC_DEBUG
2417#ifdef __FreeBSD__
2418	KASSERT(st->sync_state == PFSYNC_S_NONE,
2419		("%s: st->sync_state == PFSYNC_S_NONE", __FUNCTION__));
2420#else
2421	KASSERT(st->sync_state == PFSYNC_S_NONE);
2422#endif
2423#endif
2424
2425	if (sc->sc_len == PFSYNC_MINPKT)
2426#ifdef __FreeBSD__
2427		callout_reset(&sc->sc_tmo, 1 * hz, pfsync_timeout,
2428		    V_pfsyncif);
2429#else
2430		timeout_add_sec(&sc->sc_tmo, 1);
2431#endif
2432
2433	pfsync_q_ins(st, PFSYNC_S_INS);
2434
2435	if (ISSET(st->state_flags, PFSTATE_ACK))
2436#ifdef __FreeBSD__
2437		pfsync_sendout();
2438#else
2439		schednetisr(NETISR_PFSYNC);
2440#endif
2441	else
2442		st->sync_updates = 0;
2443}
2444
2445int defer = 10;
2446
2447int
2448pfsync_defer(struct pf_state *st, struct mbuf *m)
2449{
2450#ifdef __FreeBSD__
2451	struct pfsync_softc *sc = V_pfsyncif;
2452#else
2453	struct pfsync_softc *sc = pfsyncif;
2454#endif
2455	struct pfsync_deferral *pd;
2456
2457#ifdef __FreeBSD__
2458	PF_LOCK_ASSERT();
2459#else
2460	splassert(IPL_SOFTNET);
2461#endif
2462
2463	if (sc->sc_deferred >= 128)
2464		pfsync_undefer(TAILQ_FIRST(&sc->sc_deferrals), 0);
2465
2466	pd = pool_get(&sc->sc_pool, M_NOWAIT);
2467	if (pd == NULL)
2468		return (0);
2469	sc->sc_deferred++;
2470
2471#ifdef __FreeBSD__
2472	m->m_flags |= M_SKIP_FIREWALL;
2473#else
2474	m->m_pkthdr.pf.flags |= PF_TAG_GENERATED;
2475#endif
2476	SET(st->state_flags, PFSTATE_ACK);
2477
2478	pd->pd_st = st;
2479	pd->pd_m = m;
2480
2481	TAILQ_INSERT_TAIL(&sc->sc_deferrals, pd, pd_entry);
2482#ifdef __FreeBSD__
2483	callout_init(&pd->pd_tmo, CALLOUT_MPSAFE);
2484	callout_reset(&pd->pd_tmo, defer, pfsync_defer_tmo,
2485		pd);
2486#else
2487	timeout_set(&pd->pd_tmo, pfsync_defer_tmo, pd);
2488	timeout_add(&pd->pd_tmo, defer);
2489#endif
2490
2491	return (1);
2492}
2493
2494void
2495pfsync_undefer(struct pfsync_deferral *pd, int drop)
2496{
2497#ifdef __FreeBSD__
2498	struct pfsync_softc *sc = V_pfsyncif;
2499#else
2500	struct pfsync_softc *sc = pfsyncif;
2501#endif
2502	int s;
2503
2504#ifdef __FreeBSD__
2505	PF_LOCK_ASSERT();
2506#else
2507	splassert(IPL_SOFTNET);
2508#endif
2509
2510	TAILQ_REMOVE(&sc->sc_deferrals, pd, pd_entry);
2511	sc->sc_deferred--;
2512
2513	CLR(pd->pd_st->state_flags, PFSTATE_ACK);
2514	timeout_del(&pd->pd_tmo); /* bah */
2515	if (drop)
2516		m_freem(pd->pd_m);
2517	else {
2518		s = splnet();
2519#ifdef __FreeBSD__
2520		/* XXX: use pf_defered?! */
2521		PF_UNLOCK();
2522#endif
2523		ip_output(pd->pd_m, (void *)NULL, (void *)NULL, 0,
2524		    (void *)NULL, (void *)NULL);
2525#ifdef __FreeBSD__
2526		PF_LOCK();
2527#endif
2528		splx(s);
2529	}
2530
2531	pool_put(&sc->sc_pool, pd);
2532}
2533
2534void
2535pfsync_defer_tmo(void *arg)
2536{
2537#if defined(__FreeBSD__) && defined(VIMAGE)
2538	struct pfsync_deferral *pd = arg;
2539#endif
2540	int s;
2541
2542	s = splsoftnet();
2543#ifdef __FreeBSD__
2544	CURVNET_SET(pd->pd_m->m_pkthdr.rcvif->if_vnet); /* XXX */
2545	PF_LOCK();
2546#endif
2547	pfsync_undefer(arg, 0);
2548#ifdef __FreeBSD__
2549	PF_UNLOCK();
2550	CURVNET_RESTORE();
2551#endif
2552	splx(s);
2553}
2554
2555void
2556pfsync_deferred(struct pf_state *st, int drop)
2557{
2558#ifdef __FreeBSD__
2559	struct pfsync_softc *sc = V_pfsyncif;
2560#else
2561	struct pfsync_softc *sc = pfsyncif;
2562#endif
2563	struct pfsync_deferral *pd;
2564
2565	TAILQ_FOREACH(pd, &sc->sc_deferrals, pd_entry) {
2566		 if (pd->pd_st == st) {
2567			pfsync_undefer(pd, drop);
2568			return;
2569		}
2570	}
2571
2572	panic("pfsync_send_deferred: unable to find deferred state");
2573}
2574
2575u_int pfsync_upds = 0;
2576
2577void
2578pfsync_update_state(struct pf_state *st)
2579{
2580#ifdef __FreeBSD__
2581	struct pfsync_softc *sc = V_pfsyncif;
2582#else
2583	struct pfsync_softc *sc = pfsyncif;
2584#endif
2585	int sync = 0;
2586
2587#ifdef __FreeBSD__
2588	PF_LOCK_ASSERT();
2589#else
2590	splassert(IPL_SOFTNET);
2591#endif
2592
2593	if (sc == NULL)
2594		return;
2595
2596	if (ISSET(st->state_flags, PFSTATE_ACK))
2597		pfsync_deferred(st, 0);
2598	if (ISSET(st->state_flags, PFSTATE_NOSYNC)) {
2599		if (st->sync_state != PFSYNC_S_NONE)
2600			pfsync_q_del(st);
2601		return;
2602	}
2603
2604	if (sc->sc_len == PFSYNC_MINPKT)
2605#ifdef __FreeBSD__
2606		callout_reset(&sc->sc_tmo, 1 * hz, pfsync_timeout,
2607		    V_pfsyncif);
2608#else
2609		timeout_add_sec(&sc->sc_tmo, 1);
2610#endif
2611
2612	switch (st->sync_state) {
2613	case PFSYNC_S_UPD_C:
2614	case PFSYNC_S_UPD:
2615	case PFSYNC_S_INS:
2616		/* we're already handling it */
2617
2618		if (st->key[PF_SK_WIRE]->proto == IPPROTO_TCP) {
2619			st->sync_updates++;
2620			if (st->sync_updates >= sc->sc_maxupdates)
2621				sync = 1;
2622		}
2623		break;
2624
2625	case PFSYNC_S_IACK:
2626		pfsync_q_del(st);
2627	case PFSYNC_S_NONE:
2628		pfsync_q_ins(st, PFSYNC_S_UPD_C);
2629		st->sync_updates = 0;
2630		break;
2631
2632	default:
2633		panic("pfsync_update_state: unexpected sync state %d",
2634		    st->sync_state);
2635	}
2636
2637	if (sync || (time_uptime - st->pfsync_time) < 2) {
2638		pfsync_upds++;
2639#ifdef __FreeBSD__
2640		pfsync_sendout();
2641#else
2642		schednetisr(NETISR_PFSYNC);
2643#endif
2644	}
2645}
2646
2647void
2648pfsync_request_update(u_int32_t creatorid, u_int64_t id)
2649{
2650#ifdef __FreeBSD__
2651	struct pfsync_softc *sc = V_pfsyncif;
2652#else
2653	struct pfsync_softc *sc = pfsyncif;
2654#endif
2655	struct pfsync_upd_req_item *item;
2656	size_t nlen = sizeof(struct pfsync_upd_req);
2657	int s;
2658
2659	PF_LOCK_ASSERT();
2660
2661	/*
2662	 * this code does nothing to prevent multiple update requests for the
2663	 * same state being generated.
2664	 */
2665
2666	item = pool_get(&sc->sc_pool, PR_NOWAIT);
2667	if (item == NULL) {
2668		/* XXX stats */
2669		return;
2670	}
2671
2672	item->ur_msg.id = id;
2673	item->ur_msg.creatorid = creatorid;
2674
2675	if (TAILQ_EMPTY(&sc->sc_upd_req_list))
2676		nlen += sizeof(struct pfsync_subheader);
2677
2678#ifdef __FreeBSD__
2679	if (sc->sc_len + nlen > sc->sc_sync_if->if_mtu) {
2680#else
2681	if (sc->sc_len + nlen > sc->sc_if.if_mtu) {
2682#endif
2683		s = splnet();
2684		pfsync_sendout();
2685		splx(s);
2686
2687		nlen = sizeof(struct pfsync_subheader) +
2688		    sizeof(struct pfsync_upd_req);
2689	}
2690
2691	TAILQ_INSERT_TAIL(&sc->sc_upd_req_list, item, ur_entry);
2692	sc->sc_len += nlen;
2693
2694#ifdef __FreeBSD__
2695	pfsync_sendout();
2696#else
2697	schednetisr(NETISR_PFSYNC);
2698#endif
2699}
2700
2701void
2702pfsync_update_state_req(struct pf_state *st)
2703{
2704#ifdef __FreeBSD__
2705	struct pfsync_softc *sc = V_pfsyncif;
2706#else
2707	struct pfsync_softc *sc = pfsyncif;
2708#endif
2709
2710	PF_LOCK_ASSERT();
2711
2712	if (sc == NULL)
2713		panic("pfsync_update_state_req: nonexistant instance");
2714
2715	if (ISSET(st->state_flags, PFSTATE_NOSYNC)) {
2716		if (st->sync_state != PFSYNC_S_NONE)
2717			pfsync_q_del(st);
2718		return;
2719	}
2720
2721	switch (st->sync_state) {
2722	case PFSYNC_S_UPD_C:
2723	case PFSYNC_S_IACK:
2724		pfsync_q_del(st);
2725	case PFSYNC_S_NONE:
2726		pfsync_q_ins(st, PFSYNC_S_UPD);
2727#ifdef __FreeBSD__
2728		pfsync_sendout();
2729#else
2730		schednetisr(NETISR_PFSYNC);
2731#endif
2732		return;
2733
2734	case PFSYNC_S_INS:
2735	case PFSYNC_S_UPD:
2736	case PFSYNC_S_DEL:
2737		/* we're already handling it */
2738		return;
2739
2740	default:
2741		panic("pfsync_update_state_req: unexpected sync state %d",
2742		    st->sync_state);
2743	}
2744}
2745
2746void
2747pfsync_delete_state(struct pf_state *st)
2748{
2749#ifdef __FreeBSD__
2750	struct pfsync_softc *sc = V_pfsyncif;
2751#else
2752	struct pfsync_softc *sc = pfsyncif;
2753#endif
2754
2755#ifdef __FreeBSD__
2756	PF_LOCK_ASSERT();
2757#else
2758	splassert(IPL_SOFTNET);
2759#endif
2760
2761	if (sc == NULL)
2762		return;
2763
2764	if (ISSET(st->state_flags, PFSTATE_ACK))
2765		pfsync_deferred(st, 1);
2766	if (ISSET(st->state_flags, PFSTATE_NOSYNC)) {
2767		if (st->sync_state != PFSYNC_S_NONE)
2768			pfsync_q_del(st);
2769		return;
2770	}
2771
2772	if (sc->sc_len == PFSYNC_MINPKT)
2773#ifdef __FreeBSD__
2774		callout_reset(&sc->sc_tmo, 1 * hz, pfsync_timeout,
2775		    V_pfsyncif);
2776#else
2777		timeout_add_sec(&sc->sc_tmo, 1);
2778#endif
2779
2780	switch (st->sync_state) {
2781	case PFSYNC_S_INS:
2782		/* we never got to tell the world so just forget about it */
2783		pfsync_q_del(st);
2784		return;
2785
2786	case PFSYNC_S_UPD_C:
2787	case PFSYNC_S_UPD:
2788	case PFSYNC_S_IACK:
2789		pfsync_q_del(st);
2790		/* FALLTHROUGH to putting it on the del list */
2791
2792	case PFSYNC_S_NONE:
2793		pfsync_q_ins(st, PFSYNC_S_DEL);
2794		return;
2795
2796	default:
2797		panic("pfsync_delete_state: unexpected sync state %d",
2798		    st->sync_state);
2799	}
2800}
2801
2802void
2803pfsync_clear_states(u_int32_t creatorid, const char *ifname)
2804{
2805	struct {
2806		struct pfsync_subheader subh;
2807		struct pfsync_clr clr;
2808	} __packed r;
2809
2810#ifdef __FreeBSD__
2811	struct pfsync_softc *sc = V_pfsyncif;
2812#else
2813	struct pfsync_softc *sc = pfsyncif;
2814#endif
2815
2816#ifdef __FreeBSD__
2817	PF_LOCK_ASSERT();
2818#else
2819	splassert(IPL_SOFTNET);
2820#endif
2821
2822	if (sc == NULL)
2823		return;
2824
2825	bzero(&r, sizeof(r));
2826
2827	r.subh.action = PFSYNC_ACT_CLR;
2828	r.subh.count = htons(1);
2829
2830	strlcpy(r.clr.ifname, ifname, sizeof(r.clr.ifname));
2831	r.clr.creatorid = creatorid;
2832
2833	pfsync_send_plus(&r, sizeof(r));
2834}
2835
2836void
2837pfsync_q_ins(struct pf_state *st, int q)
2838{
2839#ifdef __FreeBSD__
2840	struct pfsync_softc *sc = V_pfsyncif;
2841#else
2842	struct pfsync_softc *sc = pfsyncif;
2843#endif
2844	size_t nlen = pfsync_qs[q].len;
2845	int s;
2846
2847	PF_LOCK_ASSERT();
2848
2849#ifdef __FreeBSD__
2850	KASSERT(st->sync_state == PFSYNC_S_NONE,
2851		("%s: st->sync_state == PFSYNC_S_NONE", __FUNCTION__));
2852#else
2853	KASSERT(st->sync_state == PFSYNC_S_NONE);
2854#endif
2855
2856#if 1 || defined(PFSYNC_DEBUG)
2857	if (sc->sc_len < PFSYNC_MINPKT)
2858#ifdef __FreeBSD__
2859		panic("pfsync pkt len is too low %zu", sc->sc_len);
2860#else
2861		panic("pfsync pkt len is too low %d", sc->sc_len);
2862#endif
2863#endif
2864	if (TAILQ_EMPTY(&sc->sc_qs[q]))
2865		nlen += sizeof(struct pfsync_subheader);
2866
2867#ifdef __FreeBSD__
2868	if (sc->sc_len + nlen > sc->sc_ifp->if_mtu) {
2869#else
2870	if (sc->sc_len + nlen > sc->sc_if.if_mtu) {
2871#endif
2872		s = splnet();
2873		pfsync_sendout();
2874		splx(s);
2875
2876		nlen = sizeof(struct pfsync_subheader) + pfsync_qs[q].len;
2877	}
2878
2879	sc->sc_len += nlen;
2880	TAILQ_INSERT_TAIL(&sc->sc_qs[q], st, sync_list);
2881	st->sync_state = q;
2882}
2883
2884void
2885pfsync_q_del(struct pf_state *st)
2886{
2887#ifdef __FreeBSD__
2888	struct pfsync_softc *sc = V_pfsyncif;
2889#else
2890	struct pfsync_softc *sc = pfsyncif;
2891#endif
2892	int q = st->sync_state;
2893
2894#ifdef __FreeBSD__
2895	KASSERT(st->sync_state != PFSYNC_S_NONE,
2896		("%s: st->sync_state != PFSYNC_S_NONE", __FUNCTION__));
2897#else
2898	KASSERT(st->sync_state != PFSYNC_S_NONE);
2899#endif
2900
2901	sc->sc_len -= pfsync_qs[q].len;
2902	TAILQ_REMOVE(&sc->sc_qs[q], st, sync_list);
2903	st->sync_state = PFSYNC_S_NONE;
2904
2905	if (TAILQ_EMPTY(&sc->sc_qs[q]))
2906		sc->sc_len -= sizeof(struct pfsync_subheader);
2907}
2908
2909#ifdef notyet
2910void
2911pfsync_update_tdb(struct tdb *t, int output)
2912{
2913#ifdef __FreeBSD__
2914	struct pfsync_softc *sc = V_pfsyncif;
2915#else
2916	struct pfsync_softc *sc = pfsyncif;
2917#endif
2918	size_t nlen = sizeof(struct pfsync_tdb);
2919	int s;
2920
2921	if (sc == NULL)
2922		return;
2923
2924	if (!ISSET(t->tdb_flags, TDBF_PFSYNC)) {
2925		if (TAILQ_EMPTY(&sc->sc_tdb_q))
2926			nlen += sizeof(struct pfsync_subheader);
2927
2928		if (sc->sc_len + nlen > sc->sc_if.if_mtu) {
2929			s = splnet();
2930			PF_LOCK();
2931			pfsync_sendout();
2932			PF_UNLOCK();
2933			splx(s);
2934
2935			nlen = sizeof(struct pfsync_subheader) +
2936			    sizeof(struct pfsync_tdb);
2937		}
2938
2939		sc->sc_len += nlen;
2940		TAILQ_INSERT_TAIL(&sc->sc_tdb_q, t, tdb_sync_entry);
2941		SET(t->tdb_flags, TDBF_PFSYNC);
2942		t->tdb_updates = 0;
2943	} else {
2944		if (++t->tdb_updates >= sc->sc_maxupdates)
2945			schednetisr(NETISR_PFSYNC);
2946	}
2947
2948	if (output)
2949		SET(t->tdb_flags, TDBF_PFSYNC_RPL);
2950	else
2951		CLR(t->tdb_flags, TDBF_PFSYNC_RPL);
2952}
2953
2954void
2955pfsync_delete_tdb(struct tdb *t)
2956{
2957#ifdef __FreeBSD__
2958	struct pfsync_softc *sc = V_pfsyncif;
2959#else
2960	struct pfsync_softc *sc = pfsyncif;
2961#endif
2962
2963	if (sc == NULL || !ISSET(t->tdb_flags, TDBF_PFSYNC))
2964		return;
2965
2966	sc->sc_len -= sizeof(struct pfsync_tdb);
2967	TAILQ_REMOVE(&sc->sc_tdb_q, t, tdb_sync_entry);
2968	CLR(t->tdb_flags, TDBF_PFSYNC);
2969
2970	if (TAILQ_EMPTY(&sc->sc_tdb_q))
2971		sc->sc_len -= sizeof(struct pfsync_subheader);
2972}
2973
2974int
2975pfsync_out_tdb(struct tdb *t, struct mbuf *m, int offset)
2976{
2977	struct pfsync_tdb *ut = (struct pfsync_tdb *)(m->m_data + offset);
2978
2979	bzero(ut, sizeof(*ut));
2980	ut->spi = t->tdb_spi;
2981	bcopy(&t->tdb_dst, &ut->dst, sizeof(ut->dst));
2982	/*
2983	 * When a failover happens, the master's rpl is probably above
2984	 * what we see here (we may be up to a second late), so
2985	 * increase it a bit for outbound tdbs to manage most such
2986	 * situations.
2987	 *
2988	 * For now, just add an offset that is likely to be larger
2989	 * than the number of packets we can see in one second. The RFC
2990	 * just says the next packet must have a higher seq value.
2991	 *
2992	 * XXX What is a good algorithm for this? We could use
2993	 * a rate-determined increase, but to know it, we would have
2994	 * to extend struct tdb.
2995	 * XXX pt->rpl can wrap over MAXINT, but if so the real tdb
2996	 * will soon be replaced anyway. For now, just don't handle
2997	 * this edge case.
2998	 */
2999#define RPL_INCR 16384
3000	ut->rpl = htonl(t->tdb_rpl + (ISSET(t->tdb_flags, TDBF_PFSYNC_RPL) ?
3001	    RPL_INCR : 0));
3002	ut->cur_bytes = htobe64(t->tdb_cur_bytes);
3003	ut->sproto = t->tdb_sproto;
3004
3005	return (sizeof(*ut));
3006}
3007#endif
3008
3009void
3010pfsync_bulk_start(void)
3011{
3012#ifdef __FreeBSD__
3013	struct pfsync_softc *sc = V_pfsyncif;
3014#else
3015	struct pfsync_softc *sc = pfsyncif;
3016#endif
3017
3018#ifdef __FreeBSD__
3019	if (V_pf_status.debug >= PF_DEBUG_MISC)
3020#else
3021	if (pf_status.debug >= PF_DEBUG_MISC)
3022#endif
3023		printf("pfsync: received bulk update request\n");
3024
3025#ifdef __FreeBSD__
3026	PF_LOCK();
3027	if (TAILQ_EMPTY(&V_state_list))
3028#else
3029	if (TAILQ_EMPTY(&state_list))
3030#endif
3031		pfsync_bulk_status(PFSYNC_BUS_END);
3032	else {
3033		sc->sc_ureq_received = time_uptime;
3034		if (sc->sc_bulk_next == NULL)
3035#ifdef __FreeBSD__
3036			sc->sc_bulk_next = TAILQ_FIRST(&V_state_list);
3037#else
3038			sc->sc_bulk_next = TAILQ_FIRST(&state_list);
3039#endif
3040			sc->sc_bulk_last = sc->sc_bulk_next;
3041
3042			pfsync_bulk_status(PFSYNC_BUS_START);
3043			callout_reset(&sc->sc_bulk_tmo, 1,
3044			    pfsync_bulk_update, sc);
3045	}
3046#ifdef __FreeBSD__
3047	PF_UNLOCK();
3048#endif
3049}
3050
3051void
3052pfsync_bulk_update(void *arg)
3053{
3054	struct pfsync_softc *sc = arg;
3055	struct pf_state *st = sc->sc_bulk_next;
3056	int i = 0;
3057	int s;
3058
3059	PF_LOCK_ASSERT();
3060
3061	s = splsoftnet();
3062#ifdef __FreeBSD__
3063	CURVNET_SET(sc->sc_ifp->if_vnet);
3064#endif
3065	for (;;) {
3066		if (st->sync_state == PFSYNC_S_NONE &&
3067		    st->timeout < PFTM_MAX &&
3068		    st->pfsync_time <= sc->sc_ureq_received) {
3069			pfsync_update_state_req(st);
3070			i++;
3071		}
3072
3073		st = TAILQ_NEXT(st, entry_list);
3074		if (st == NULL)
3075#ifdef __FreeBSD__
3076			st = TAILQ_FIRST(&V_state_list);
3077#else
3078			st = TAILQ_FIRST(&state_list);
3079#endif
3080
3081		if (st == sc->sc_bulk_last) {
3082			/* we're done */
3083			sc->sc_bulk_next = NULL;
3084			sc->sc_bulk_last = NULL;
3085			pfsync_bulk_status(PFSYNC_BUS_END);
3086			break;
3087		}
3088
3089#ifdef __FreeBSD__
3090		if (i > 1 && (sc->sc_ifp->if_mtu - sc->sc_len) <
3091#else
3092		if (i > 1 && (sc->sc_if.if_mtu - sc->sc_len) <
3093#endif
3094		    sizeof(struct pfsync_state)) {
3095			/* we've filled a packet */
3096			sc->sc_bulk_next = st;
3097#ifdef __FreeBSD__
3098			callout_reset(&sc->sc_bulk_tmo, 1,
3099			    pfsync_bulk_update, sc);
3100#else
3101			timeout_add(&sc->sc_bulk_tmo, 1);
3102#endif
3103			break;
3104		}
3105	}
3106
3107#ifdef __FreeBSD__
3108	CURVNET_RESTORE();
3109#endif
3110	splx(s);
3111}
3112
3113void
3114pfsync_bulk_status(u_int8_t status)
3115{
3116	struct {
3117		struct pfsync_subheader subh;
3118		struct pfsync_bus bus;
3119	} __packed r;
3120
3121#ifdef __FreeBSD__
3122	struct pfsync_softc *sc = V_pfsyncif;
3123#else
3124	struct pfsync_softc *sc = pfsyncif;
3125#endif
3126
3127	PF_LOCK_ASSERT();
3128
3129	bzero(&r, sizeof(r));
3130
3131	r.subh.action = PFSYNC_ACT_BUS;
3132	r.subh.count = htons(1);
3133
3134#ifdef __FreeBSD__
3135	r.bus.creatorid = V_pf_status.hostid;
3136#else
3137	r.bus.creatorid = pf_status.hostid;
3138#endif
3139	r.bus.endtime = htonl(time_uptime - sc->sc_ureq_received);
3140	r.bus.status = status;
3141
3142	pfsync_send_plus(&r, sizeof(r));
3143}
3144
3145void
3146pfsync_bulk_fail(void *arg)
3147{
3148	struct pfsync_softc *sc = arg;
3149
3150#ifdef __FreeBSD__
3151	CURVNET_SET(sc->sc_ifp->if_vnet);
3152#endif
3153
3154	if (sc->sc_bulk_tries++ < PFSYNC_MAX_BULKTRIES) {
3155		/* Try again */
3156#ifdef __FreeBSD__
3157		callout_reset(&sc->sc_bulkfail_tmo, 5 * hz,
3158		    pfsync_bulk_fail, V_pfsyncif);
3159#else
3160		timeout_add_sec(&sc->sc_bulkfail_tmo, 5);
3161#endif
3162		PF_LOCK();
3163		pfsync_request_update(0, 0);
3164		PF_UNLOCK();
3165	} else {
3166		/* Pretend like the transfer was ok */
3167		sc->sc_ureq_sent = 0;
3168		sc->sc_bulk_tries = 0;
3169#if NCARP > 0
3170#ifdef notyet
3171#ifdef __FreeBSD__
3172		if (!sc->pfsync_sync_ok)
3173#else
3174		if (!pfsync_sync_ok)
3175#endif
3176			carp_group_demote_adj(&sc->sc_if, -1);
3177#endif
3178#endif
3179#ifdef __FreeBSD__
3180		sc->pfsync_sync_ok = 1;
3181#else
3182		pfsync_sync_ok = 1;
3183#endif
3184#ifdef __FreeBSD__
3185		if (V_pf_status.debug >= PF_DEBUG_MISC)
3186#else
3187		if (pf_status.debug >= PF_DEBUG_MISC)
3188#endif
3189			printf("pfsync: failed to receive bulk update\n");
3190	}
3191
3192#ifdef __FreeBSD__
3193	CURVNET_RESTORE();
3194#endif
3195}
3196
3197void
3198pfsync_send_plus(void *plus, size_t pluslen)
3199{
3200#ifdef __FreeBSD__
3201	struct pfsync_softc *sc = V_pfsyncif;
3202#else
3203	struct pfsync_softc *sc = pfsyncif;
3204#endif
3205	int s;
3206
3207	PF_LOCK_ASSERT();
3208
3209#ifdef __FreeBSD__
3210	if (sc->sc_len + pluslen > sc->sc_ifp->if_mtu) {
3211#else
3212	if (sc->sc_len + pluslen > sc->sc_if.if_mtu) {
3213#endif
3214		s = splnet();
3215		pfsync_sendout();
3216		splx(s);
3217	}
3218
3219	sc->sc_plus = plus;
3220	sc->sc_len += (sc->sc_pluslen = pluslen);
3221
3222	s = splnet();
3223	pfsync_sendout();
3224	splx(s);
3225}
3226
3227int
3228pfsync_up(void)
3229{
3230#ifdef __FreeBSD__
3231	struct pfsync_softc *sc = V_pfsyncif;
3232#else
3233	struct pfsync_softc *sc = pfsyncif;
3234#endif
3235
3236#ifdef __FreeBSD__
3237	if (sc == NULL || !ISSET(sc->sc_ifp->if_flags, IFF_DRV_RUNNING))
3238#else
3239	if (sc == NULL || !ISSET(sc->sc_if.if_flags, IFF_RUNNING))
3240#endif
3241		return (0);
3242
3243	return (1);
3244}
3245
3246int
3247pfsync_state_in_use(struct pf_state *st)
3248{
3249#ifdef __FreeBSD__
3250	struct pfsync_softc *sc = V_pfsyncif;
3251#else
3252	struct pfsync_softc *sc = pfsyncif;
3253#endif
3254
3255	if (sc == NULL)
3256		return (0);
3257
3258	if (st->sync_state != PFSYNC_S_NONE ||
3259	    st == sc->sc_bulk_next ||
3260	    st == sc->sc_bulk_last)
3261		return (1);
3262
3263	return (0);
3264}
3265
3266u_int pfsync_ints;
3267u_int pfsync_tmos;
3268
3269void
3270pfsync_timeout(void *arg)
3271{
3272#if defined(__FreeBSD__) && defined(VIMAGE)
3273	struct pfsync_softc *sc = arg;
3274#endif
3275	int s;
3276
3277#ifdef __FreeBSD__
3278	CURVNET_SET(sc->sc_ifp->if_vnet);
3279#endif
3280
3281	pfsync_tmos++;
3282
3283	s = splnet();
3284#ifdef __FreeBSD__
3285	PF_LOCK();
3286#endif
3287	pfsync_sendout();
3288#ifdef __FreeBSD__
3289	PF_UNLOCK();
3290#endif
3291	splx(s);
3292
3293#ifdef __FreeBSD__
3294	CURVNET_RESTORE();
3295#endif
3296}
3297
3298/* this is a softnet/netisr handler */
3299void
3300#ifdef __FreeBSD__
3301pfsyncintr(void *arg)
3302{
3303	struct pfsync_softc *sc = arg;
3304	struct mbuf *m, *n;
3305
3306	CURVNET_SET(sc->sc_ifp->if_vnet);
3307	pfsync_ints++;
3308
3309	IF_DEQUEUE_ALL(&sc->sc_ifp->if_snd, m);
3310
3311	for (; m != NULL; m = n) {
3312
3313		n = m->m_nextpkt;
3314		m->m_nextpkt = NULL;
3315		if (ip_output(m, NULL, NULL, IP_RAWOUTPUT, &sc->sc_imo, NULL)
3316		    == 0)
3317			V_pfsyncstats.pfsyncs_opackets++;
3318		else
3319			V_pfsyncstats.pfsyncs_oerrors++;
3320	}
3321	CURVNET_RESTORE();
3322}
3323#else
3324pfsyncintr(void)
3325{
3326	int s;
3327
3328	pfsync_ints++;
3329
3330	s = splnet();
3331	pfsync_sendout();
3332	splx(s);
3333}
3334#endif
3335
3336int
3337pfsync_sysctl(int *name, u_int namelen, void *oldp, size_t *oldlenp, void *newp,
3338    size_t newlen)
3339{
3340
3341#ifdef notyet
3342	/* All sysctl names at this level are terminal. */
3343	if (namelen != 1)
3344		return (ENOTDIR);
3345
3346	switch (name[0]) {
3347	case PFSYNCCTL_STATS:
3348		if (newp != NULL)
3349			return (EPERM);
3350		return (sysctl_struct(oldp, oldlenp, newp, newlen,
3351		    &V_pfsyncstats, sizeof(V_pfsyncstats)));
3352	}
3353#endif
3354	return (ENOPROTOOPT);
3355}
3356
3357#ifdef __FreeBSD__
3358void
3359pfsync_ifdetach(void *arg, struct ifnet *ifp)
3360{
3361	struct pfsync_softc *sc = (struct pfsync_softc *)arg;
3362	struct ip_moptions *imo;
3363
3364	if (sc == NULL || sc->sc_sync_if != ifp)
3365		return;         /* not for us; unlocked read */
3366
3367	CURVNET_SET(sc->sc_ifp->if_vnet);
3368
3369	PF_LOCK();
3370
3371	/* Deal with a member interface going away from under us. */
3372	sc->sc_sync_if = NULL;
3373	imo = &sc->sc_imo;
3374	if (imo->imo_num_memberships > 0) {
3375		KASSERT(imo->imo_num_memberships == 1,
3376		    ("%s: imo_num_memberships != 1", __func__));
3377		/*
3378		 * Our event handler is always called after protocol
3379		 * domains have been detached from the underlying ifnet.
3380		 * Do not call in_delmulti(); we held a single reference
3381		 * which the protocol domain has purged in in_purgemaddrs().
3382		 */
3383		PF_UNLOCK();
3384		imo->imo_membership[--imo->imo_num_memberships] = NULL;
3385		PF_LOCK();
3386		imo->imo_multicast_ifp = NULL;
3387	}
3388
3389	PF_UNLOCK();
3390
3391	CURVNET_RESTORE();
3392}
3393
3394static int
3395vnet_pfsync_init(const void *unused)
3396{
3397	int error = 0;
3398
3399	pfsyncattach(0);
3400
3401	error = swi_add(NULL, "pfsync", pfsyncintr, V_pfsyncif,
3402		SWI_NET, INTR_MPSAFE, &pfsync_swi.pfsync_swi_cookie);
3403	if (error)
3404		panic("%s: swi_add %d", __func__, error);
3405
3406	PF_LOCK();
3407	pfsync_state_import_ptr = pfsync_state_import;
3408	pfsync_up_ptr = pfsync_up;
3409	pfsync_insert_state_ptr = pfsync_insert_state;
3410	pfsync_update_state_ptr = pfsync_update_state;
3411	pfsync_delete_state_ptr = pfsync_delete_state;
3412	pfsync_clear_states_ptr = pfsync_clear_states;
3413	pfsync_state_in_use_ptr = pfsync_state_in_use;
3414	pfsync_defer_ptr = pfsync_defer;
3415	PF_UNLOCK();
3416
3417	return (0);
3418}
3419
3420static int
3421vnet_pfsync_uninit(const void *unused)
3422{
3423
3424	swi_remove(pfsync_swi.pfsync_swi_cookie);
3425
3426	PF_LOCK();
3427	pfsync_state_import_ptr = NULL;
3428	pfsync_up_ptr = NULL;
3429	pfsync_insert_state_ptr = NULL;
3430	pfsync_update_state_ptr = NULL;
3431	pfsync_delete_state_ptr = NULL;
3432	pfsync_clear_states_ptr = NULL;
3433	pfsync_state_in_use_ptr = NULL;
3434	pfsync_defer_ptr = NULL;
3435	PF_UNLOCK();
3436
3437	if_clone_detach(&pfsync_cloner);
3438
3439	return (0);
3440}
3441
3442/* Define startup order. */
3443#define	PFSYNC_SYSINIT_ORDER	SI_SUB_PROTO_IF
3444#define	PFSYNC_MODEVENT_ORDER	(SI_ORDER_FIRST) /* On boot slot in here. */
3445#define	PFSYNC_VNET_ORDER	(PFSYNC_MODEVENT_ORDER + 2) /* Later still. */
3446
3447/*
3448 * Starting up.
3449 * VNET_SYSINIT is called for each existing vnet and each new vnet.
3450 */
3451VNET_SYSINIT(vnet_pfsync_init, PFSYNC_SYSINIT_ORDER, PFSYNC_VNET_ORDER,
3452    vnet_pfsync_init, NULL);
3453
3454/*
3455 * Closing up shop. These are done in REVERSE ORDER,
3456 * Not called on reboot.
3457 * VNET_SYSUNINIT is called for each exiting vnet as it exits.
3458 */
3459VNET_SYSUNINIT(vnet_pfsync_uninit, PFSYNC_SYSINIT_ORDER, PFSYNC_VNET_ORDER,
3460    vnet_pfsync_uninit, NULL);
3461static int
3462pfsync_modevent(module_t mod, int type, void *data)
3463{
3464	int error = 0;
3465
3466	switch (type) {
3467	case MOD_LOAD:
3468#ifndef __FreeBSD__
3469		pfsyncattach(0);
3470#endif
3471		break;
3472	case MOD_UNLOAD:
3473#ifndef __FreeBSD__
3474		if_clone_detach(&pfsync_cloner);
3475#endif
3476		break;
3477	default:
3478		error = EINVAL;
3479		break;
3480	}
3481
3482	return error;
3483}
3484
3485static moduledata_t pfsync_mod = {
3486	"pfsync",
3487	pfsync_modevent,
3488	0
3489};
3490
3491#define PFSYNC_MODVER 1
3492
3493DECLARE_MODULE(pfsync, pfsync_mod, SI_SUB_PSEUDO, SI_ORDER_ANY);
3494MODULE_VERSION(pfsync, PFSYNC_MODVER);
3495MODULE_DEPEND(pfsync, pf, PF_MODVER, PF_MODVER, PF_MODVER);
3496#endif /* __FreeBSD__ */
3497