if_pfsync.c revision 228732
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#ifdef __FreeBSD__
46#include "opt_inet.h"
47#include "opt_inet6.h"
48#include "opt_bpf.h"
49#include "opt_pf.h"
50
51#include <sys/cdefs.h>
52__FBSDID("$FreeBSD: head/sys/contrib/pf/net/if_pfsync.c 228732 2011-12-20 12:34:16Z glebius $");
53
54#ifdef DEV_BPF
55#define	NBPFILTER	DEV_BPF
56#else
57#define	NBPFILTER	0
58#endif
59
60#ifdef DEV_PFSYNC
61#define	NPFSYNC		DEV_PFSYNC
62#else
63#define	NPFSYNC		0
64#endif
65
66#ifdef DEV_CARP
67#define	NCARP		DEV_CARP
68#else
69#define	NCARP		0
70#endif
71#endif /* __FreeBSD__ */
72
73#include <sys/param.h>
74#include <sys/kernel.h>
75#ifdef __FreeBSD__
76#include <sys/bus.h>
77#include <sys/interrupt.h>
78#include <sys/priv.h>
79#endif
80#include <sys/proc.h>
81#include <sys/systm.h>
82#include <sys/time.h>
83#include <sys/mbuf.h>
84#include <sys/socket.h>
85#ifdef __FreeBSD__
86#include <sys/endian.h>
87#include <sys/malloc.h>
88#include <sys/module.h>
89#include <sys/sockio.h>
90#include <sys/taskqueue.h>
91#include <sys/lock.h>
92#include <sys/mutex.h>
93#else
94#include <sys/ioctl.h>
95#include <sys/timeout.h>
96#endif
97#include <sys/sysctl.h>
98#ifndef __FreeBSD__
99#include <sys/pool.h>
100#endif
101
102#include <net/if.h>
103#ifdef __FreeBSD__
104#include <net/if_clone.h>
105#endif
106#include <net/if_types.h>
107#include <net/route.h>
108#include <net/bpf.h>
109#include <net/netisr.h>
110#ifdef __FreeBSD__
111#include <net/vnet.h>
112#endif
113
114#include <netinet/in.h>
115#include <netinet/if_ether.h>
116#include <netinet/tcp.h>
117#include <netinet/tcp_seq.h>
118
119#ifdef	INET
120#include <netinet/in_systm.h>
121#include <netinet/in_var.h>
122#include <netinet/ip.h>
123#include <netinet/ip_var.h>
124#endif
125
126#ifdef INET6
127#include <netinet6/nd6.h>
128#endif /* INET6 */
129
130#ifndef __FreeBSD__
131#include "carp.h"
132#endif
133#if NCARP > 0
134#include <netinet/ip_carp.h>
135#endif
136
137#include <net/pfvar.h>
138#include <net/if_pfsync.h>
139
140#ifndef __FreeBSD__
141#include "bpfilter.h"
142#include "pfsync.h"
143#endif
144
145#define PFSYNC_MINPKT ( \
146	sizeof(struct ip) + \
147	sizeof(struct pfsync_header) + \
148	sizeof(struct pfsync_subheader) + \
149	sizeof(struct pfsync_eof))
150
151struct pfsync_pkt {
152	struct ip *ip;
153	struct in_addr src;
154	u_int8_t flags;
155};
156
157int	pfsync_input_hmac(struct mbuf *, int);
158
159int	pfsync_upd_tcp(struct pf_state *, struct pfsync_state_peer *,
160	    struct pfsync_state_peer *);
161
162int	pfsync_in_clr(struct pfsync_pkt *, struct mbuf *, int, int);
163int	pfsync_in_ins(struct pfsync_pkt *, struct mbuf *, int, int);
164int	pfsync_in_iack(struct pfsync_pkt *, struct mbuf *, int, int);
165int	pfsync_in_upd(struct pfsync_pkt *, struct mbuf *, int, int);
166int	pfsync_in_upd_c(struct pfsync_pkt *, struct mbuf *, int, int);
167int	pfsync_in_ureq(struct pfsync_pkt *, struct mbuf *, int, int);
168int	pfsync_in_del(struct pfsync_pkt *, struct mbuf *, int, int);
169int	pfsync_in_del_c(struct pfsync_pkt *, struct mbuf *, int, int);
170int	pfsync_in_bus(struct pfsync_pkt *, struct mbuf *, int, int);
171int	pfsync_in_tdb(struct pfsync_pkt *, struct mbuf *, int, int);
172int	pfsync_in_eof(struct pfsync_pkt *, struct mbuf *, int, int);
173
174int	pfsync_in_error(struct pfsync_pkt *, struct mbuf *, int, int);
175
176int	(*pfsync_acts[])(struct pfsync_pkt *, struct mbuf *, int, int) = {
177	pfsync_in_clr,			/* PFSYNC_ACT_CLR */
178	pfsync_in_ins,			/* PFSYNC_ACT_INS */
179	pfsync_in_iack,			/* PFSYNC_ACT_INS_ACK */
180	pfsync_in_upd,			/* PFSYNC_ACT_UPD */
181	pfsync_in_upd_c,		/* PFSYNC_ACT_UPD_C */
182	pfsync_in_ureq,			/* PFSYNC_ACT_UPD_REQ */
183	pfsync_in_del,			/* PFSYNC_ACT_DEL */
184	pfsync_in_del_c,		/* PFSYNC_ACT_DEL_C */
185	pfsync_in_error,		/* PFSYNC_ACT_INS_F */
186	pfsync_in_error,		/* PFSYNC_ACT_DEL_F */
187	pfsync_in_bus,			/* PFSYNC_ACT_BUS */
188	pfsync_in_tdb,			/* PFSYNC_ACT_TDB */
189	pfsync_in_eof			/* PFSYNC_ACT_EOF */
190};
191
192struct pfsync_q {
193	int		(*write)(struct pf_state *, struct mbuf *, int);
194	size_t		len;
195	u_int8_t	action;
196};
197
198/* we have one of these for every PFSYNC_S_ */
199int	pfsync_out_state(struct pf_state *, struct mbuf *, int);
200int	pfsync_out_iack(struct pf_state *, struct mbuf *, int);
201int	pfsync_out_upd_c(struct pf_state *, struct mbuf *, int);
202int	pfsync_out_del(struct pf_state *, struct mbuf *, int);
203
204struct pfsync_q pfsync_qs[] = {
205	{ pfsync_out_state, sizeof(struct pfsync_state),   PFSYNC_ACT_INS },
206	{ pfsync_out_iack,  sizeof(struct pfsync_ins_ack), PFSYNC_ACT_INS_ACK },
207	{ pfsync_out_state, sizeof(struct pfsync_state),   PFSYNC_ACT_UPD },
208	{ pfsync_out_upd_c, sizeof(struct pfsync_upd_c),   PFSYNC_ACT_UPD_C },
209	{ pfsync_out_del,   sizeof(struct pfsync_del_c),   PFSYNC_ACT_DEL_C }
210};
211
212void	pfsync_q_ins(struct pf_state *, int);
213void	pfsync_q_del(struct pf_state *);
214
215struct pfsync_upd_req_item {
216	TAILQ_ENTRY(pfsync_upd_req_item)	ur_entry;
217	struct pfsync_upd_req			ur_msg;
218};
219TAILQ_HEAD(pfsync_upd_reqs, pfsync_upd_req_item);
220
221struct pfsync_deferral {
222	TAILQ_ENTRY(pfsync_deferral)		 pd_entry;
223	struct pf_state				*pd_st;
224	struct mbuf				*pd_m;
225#ifdef __FreeBSD__
226	struct callout				 pd_tmo;
227#else
228	struct timeout				 pd_tmo;
229#endif
230};
231TAILQ_HEAD(pfsync_deferrals, pfsync_deferral);
232
233#define PFSYNC_PLSIZE	MAX(sizeof(struct pfsync_upd_req_item), \
234			    sizeof(struct pfsync_deferral))
235
236#ifdef notyet
237int	pfsync_out_tdb(struct tdb *, struct mbuf *, int);
238#endif
239
240struct pfsync_softc {
241#ifdef __FreeBSD__
242	struct ifnet		*sc_ifp;
243#else
244	struct ifnet		 sc_if;
245#endif
246	struct ifnet		*sc_sync_if;
247
248#ifdef __FreeBSD__
249	uma_zone_t		 sc_pool;
250#else
251	struct pool		 sc_pool;
252#endif
253
254	struct ip_moptions	 sc_imo;
255
256	struct in_addr		 sc_sync_peer;
257	u_int8_t		 sc_maxupdates;
258#ifdef __FreeBSD__
259	int			 pfsync_sync_ok;
260#endif
261
262	struct ip		 sc_template;
263
264	struct pf_state_queue	 sc_qs[PFSYNC_S_COUNT];
265	size_t			 sc_len;
266
267	struct pfsync_upd_reqs	 sc_upd_req_list;
268
269	struct pfsync_deferrals	 sc_deferrals;
270	u_int			 sc_deferred;
271
272	void			*sc_plus;
273	size_t			 sc_pluslen;
274
275	u_int32_t		 sc_ureq_sent;
276	int			 sc_bulk_tries;
277#ifdef __FreeBSD__
278	struct callout		 sc_bulkfail_tmo;
279#else
280	struct timeout		 sc_bulkfail_tmo;
281#endif
282
283	u_int32_t		 sc_ureq_received;
284	struct pf_state		*sc_bulk_next;
285	struct pf_state		*sc_bulk_last;
286#ifdef __FreeBSD__
287	struct callout		 sc_bulk_tmo;
288#else
289	struct timeout		 sc_bulk_tmo;
290#endif
291
292	TAILQ_HEAD(, tdb)	 sc_tdb_q;
293
294#ifdef __FreeBSD__
295	struct callout		 sc_tmo;
296#else
297	struct timeout		 sc_tmo;
298#endif
299#ifdef __FreeBSD__
300	eventhandler_tag	 sc_detachtag;
301#endif
302
303};
304
305#ifdef __FreeBSD__
306static VNET_DEFINE(struct pfsync_softc	*, pfsyncif) = NULL;
307#define	V_pfsyncif		VNET(pfsyncif)
308
309static VNET_DEFINE(struct pfsyncstats, pfsyncstats);
310#define	V_pfsyncstats		VNET(pfsyncstats)
311
312SYSCTL_NODE(_net, OID_AUTO, pfsync, CTLFLAG_RW, 0, "PFSYNC");
313SYSCTL_VNET_STRUCT(_net_pfsync, OID_AUTO, stats, CTLFLAG_RW,
314    &VNET_NAME(pfsyncstats), pfsyncstats,
315    "PFSYNC statistics (struct pfsyncstats, net/if_pfsync.h)");
316#else
317struct pfsync_softc	*pfsyncif = NULL;
318struct pfsyncstats	 pfsyncstats;
319#define	V_pfsyncstats	 pfsyncstats
320#endif
321
322#ifdef __FreeBSD__
323static void	pfsyncintr(void *);
324struct pfsync_swi {
325	void *	pfsync_swi_cookie;
326};
327static struct pfsync_swi	 pfsync_swi;
328#define	schednetisr(p)	swi_sched(pfsync_swi.pfsync_swi_cookie, 0)
329#define	NETISR_PFSYNC
330#endif
331
332void	pfsyncattach(int);
333#ifdef __FreeBSD__
334int	pfsync_clone_create(struct if_clone *, int, caddr_t);
335void	pfsync_clone_destroy(struct ifnet *);
336#else
337int	pfsync_clone_create(struct if_clone *, int);
338int	pfsync_clone_destroy(struct ifnet *);
339#endif
340int	pfsync_alloc_scrub_memory(struct pfsync_state_peer *,
341	    struct pf_state_peer *);
342void	pfsync_update_net_tdb(struct pfsync_tdb *);
343int	pfsyncoutput(struct ifnet *, struct mbuf *, struct sockaddr *,
344#ifdef __FreeBSD__
345	    struct route *);
346#else
347	    struct rtentry *);
348#endif
349int	pfsyncioctl(struct ifnet *, u_long, caddr_t);
350void	pfsyncstart(struct ifnet *);
351
352struct mbuf *pfsync_if_dequeue(struct ifnet *);
353struct mbuf *pfsync_get_mbuf(struct pfsync_softc *);
354
355void	pfsync_deferred(struct pf_state *, int);
356void	pfsync_undefer(struct pfsync_deferral *, int);
357void	pfsync_defer_tmo(void *);
358
359void	pfsync_request_update(u_int32_t, u_int64_t);
360void	pfsync_update_state_req(struct pf_state *);
361
362void	pfsync_drop(struct pfsync_softc *);
363void	pfsync_sendout(void);
364void	pfsync_send_plus(void *, size_t);
365int	pfsync_tdb_sendout(struct pfsync_softc *);
366int	pfsync_sendout_mbuf(struct pfsync_softc *, struct mbuf *);
367void	pfsync_timeout(void *);
368void	pfsync_tdb_timeout(void *);
369void	pfsync_send_bus(struct pfsync_softc *, u_int8_t);
370
371void	pfsync_bulk_start(void);
372void	pfsync_bulk_status(u_int8_t);
373void	pfsync_bulk_update(void *);
374void	pfsync_bulk_fail(void *);
375
376#ifdef __FreeBSD__
377void	pfsync_ifdetach(void *, struct ifnet *);
378
379/* XXX: ugly */
380#define	betoh64		(unsigned long long)be64toh
381#define	timeout_del	callout_stop
382#endif
383
384#define PFSYNC_MAX_BULKTRIES	12
385#ifndef __FreeBSD__
386int	pfsync_sync_ok;
387#endif
388
389#ifdef __FreeBSD__
390IFC_SIMPLE_DECLARE(pfsync, 1);
391#else
392struct if_clone	pfsync_cloner =
393    IF_CLONE_INITIALIZER("pfsync", pfsync_clone_create, pfsync_clone_destroy);
394#endif
395
396void
397pfsyncattach(int npfsync)
398{
399	if_clone_attach(&pfsync_cloner);
400}
401int
402#ifdef __FreeBSD__
403pfsync_clone_create(struct if_clone *ifc, int unit, caddr_t param)
404#else
405pfsync_clone_create(struct if_clone *ifc, int unit)
406#endif
407{
408	struct pfsync_softc *sc;
409	struct ifnet *ifp;
410	int q;
411
412	if (unit != 0)
413		return (EINVAL);
414
415#ifndef __FreeBSD__
416	pfsync_sync_ok = 1;
417#endif
418
419	sc = malloc(sizeof(struct pfsync_softc), M_DEVBUF, M_NOWAIT | M_ZERO);
420	if (sc == NULL)
421		return (ENOMEM);
422
423	for (q = 0; q < PFSYNC_S_COUNT; q++)
424		TAILQ_INIT(&sc->sc_qs[q]);
425
426#ifdef __FreeBSD__
427	sc->pfsync_sync_ok = 1;
428	sc->sc_pool = uma_zcreate("pfsync", PFSYNC_PLSIZE,
429			NULL, NULL, NULL, NULL, UMA_ALIGN_PTR, 0);
430	if (sc->sc_pool == NULL) {
431		free(sc, M_DEVBUF);
432		return (ENOMEM);
433	}
434#else
435	pool_init(&sc->sc_pool, PFSYNC_PLSIZE, 0, 0, 0, "pfsync", NULL);
436#endif
437	TAILQ_INIT(&sc->sc_upd_req_list);
438	TAILQ_INIT(&sc->sc_deferrals);
439	sc->sc_deferred = 0;
440
441	TAILQ_INIT(&sc->sc_tdb_q);
442
443	sc->sc_len = PFSYNC_MINPKT;
444	sc->sc_maxupdates = 128;
445
446#ifdef __FreeBSD__
447	sc->sc_imo.imo_membership = (struct in_multi **)malloc(
448	    (sizeof(struct in_multi *) * IP_MIN_MEMBERSHIPS), M_DEVBUF,
449	    M_NOWAIT | M_ZERO);
450	sc->sc_imo.imo_max_memberships = IP_MIN_MEMBERSHIPS;
451	sc->sc_imo.imo_multicast_vif = -1;
452#else
453	sc->sc_imo.imo_membership = (struct in_multi **)malloc(
454	    (sizeof(struct in_multi *) * IP_MIN_MEMBERSHIPS), M_IPMOPTS,
455	    M_WAITOK | M_ZERO);
456	sc->sc_imo.imo_max_memberships = IP_MIN_MEMBERSHIPS;
457#endif
458
459#ifdef __FreeBSD__
460	ifp = sc->sc_ifp = if_alloc(IFT_PFSYNC);
461	if (ifp == NULL) {
462		free(sc->sc_imo.imo_membership, M_DEVBUF);
463		uma_zdestroy(sc->sc_pool);
464		free(sc, M_DEVBUF);
465		return (ENOSPC);
466	}
467	if_initname(ifp, ifc->ifc_name, unit);
468
469	sc->sc_detachtag = EVENTHANDLER_REGISTER(ifnet_departure_event,
470#ifdef __FreeBSD__
471	    pfsync_ifdetach, V_pfsyncif, EVENTHANDLER_PRI_ANY);
472#else
473	    pfsync_ifdetach, pfsyncif, EVENTHANDLER_PRI_ANY);
474#endif
475	if (sc->sc_detachtag == NULL) {
476		if_free(ifp);
477		free(sc->sc_imo.imo_membership, M_DEVBUF);
478		uma_zdestroy(sc->sc_pool);
479		free(sc, M_DEVBUF);
480		return (ENOSPC);
481	}
482#else
483	ifp = &sc->sc_if;
484	snprintf(ifp->if_xname, sizeof ifp->if_xname, "pfsync%d", unit);
485#endif
486	ifp->if_softc = sc;
487	ifp->if_ioctl = pfsyncioctl;
488	ifp->if_output = pfsyncoutput;
489	ifp->if_start = pfsyncstart;
490	ifp->if_type = IFT_PFSYNC;
491	ifp->if_snd.ifq_maxlen = ifqmaxlen;
492	ifp->if_hdrlen = sizeof(struct pfsync_header);
493	ifp->if_mtu = 1500; /* XXX */
494#ifdef __FreeBSD__
495	callout_init(&sc->sc_tmo, CALLOUT_MPSAFE);
496	callout_init_mtx(&sc->sc_bulk_tmo, &pf_task_mtx, 0);
497	callout_init(&sc->sc_bulkfail_tmo, CALLOUT_MPSAFE);
498#else
499	ifp->if_hardmtu = MCLBYTES; /* XXX */
500	timeout_set(&sc->sc_tmo, pfsync_timeout, sc);
501	timeout_set(&sc->sc_bulk_tmo, pfsync_bulk_update, sc);
502	timeout_set(&sc->sc_bulkfail_tmo, pfsync_bulk_fail, sc);
503#endif
504
505	if_attach(ifp);
506#ifndef __FreeBSD__
507	if_alloc_sadl(ifp);
508#endif
509
510#if NCARP > 0
511	if_addgroup(ifp, "carp");
512#endif
513
514#if NBPFILTER > 0
515#ifdef __FreeBSD__
516	bpfattach(ifp, DLT_PFSYNC, PFSYNC_HDRLEN);
517#else
518	bpfattach(&sc->sc_if.if_bpf, ifp, DLT_PFSYNC, PFSYNC_HDRLEN);
519#endif
520#endif
521
522#ifdef __FreeBSD__
523	V_pfsyncif = sc;
524#else
525	pfsyncif = sc;
526#endif
527
528	return (0);
529}
530
531#ifdef __FreeBSD__
532void
533#else
534int
535#endif
536pfsync_clone_destroy(struct ifnet *ifp)
537{
538	struct pfsync_softc *sc = ifp->if_softc;
539
540#ifdef __FreeBSD__
541	EVENTHANDLER_DEREGISTER(ifnet_departure_event, sc->sc_detachtag);
542	PF_LOCK();
543#endif
544	timeout_del(&sc->sc_bulk_tmo);
545	timeout_del(&sc->sc_tmo);
546#ifdef __FreeBSD__
547	PF_UNLOCK();
548#endif
549#if NCARP > 0
550#ifdef notyet
551#ifdef __FreeBSD__
552	if (!sc->pfsync_sync_ok)
553#else
554	if (!pfsync_sync_ok)
555#endif
556		carp_group_demote_adj(&sc->sc_if, -1);
557#endif
558#endif
559#if NBPFILTER > 0
560	bpfdetach(ifp);
561#endif
562	if_detach(ifp);
563
564	pfsync_drop(sc);
565
566	while (sc->sc_deferred > 0)
567		pfsync_undefer(TAILQ_FIRST(&sc->sc_deferrals), 0);
568
569#ifdef __FreeBSD__
570	UMA_DESTROY(sc->sc_pool);
571#else
572	pool_destroy(&sc->sc_pool);
573#endif
574#ifdef __FreeBSD__
575	if_free(ifp);
576	free(sc->sc_imo.imo_membership, M_DEVBUF);
577#else
578	free(sc->sc_imo.imo_membership, M_IPMOPTS);
579#endif
580	free(sc, M_DEVBUF);
581
582#ifdef __FreeBSD__
583	V_pfsyncif = NULL;
584#else
585	pfsyncif = NULL;
586#endif
587
588#ifndef __FreeBSD__
589	return (0);
590#endif
591}
592
593struct mbuf *
594pfsync_if_dequeue(struct ifnet *ifp)
595{
596	struct mbuf *m;
597#ifndef __FreeBSD__
598	int s;
599#endif
600
601#ifdef __FreeBSD__
602	IF_LOCK(&ifp->if_snd);
603	_IF_DROP(&ifp->if_snd);
604	_IF_DEQUEUE(&ifp->if_snd, m);
605	IF_UNLOCK(&ifp->if_snd);
606#else
607	s = splnet();
608	IF_DEQUEUE(&ifp->if_snd, m);
609	splx(s);
610#endif
611
612	return (m);
613}
614
615/*
616 * Start output on the pfsync interface.
617 */
618void
619pfsyncstart(struct ifnet *ifp)
620{
621	struct mbuf *m;
622
623	while ((m = pfsync_if_dequeue(ifp)) != NULL) {
624#ifndef __FreeBSD__
625		IF_DROP(&ifp->if_snd);
626#endif
627		m_freem(m);
628	}
629}
630
631int
632pfsync_alloc_scrub_memory(struct pfsync_state_peer *s,
633    struct pf_state_peer *d)
634{
635	if (s->scrub.scrub_flag && d->scrub == NULL) {
636#ifdef __FreeBSD__
637		d->scrub = pool_get(&V_pf_state_scrub_pl, PR_NOWAIT | PR_ZERO);
638#else
639		d->scrub = pool_get(&pf_state_scrub_pl, PR_NOWAIT | PR_ZERO);
640#endif
641		if (d->scrub == NULL)
642			return (ENOMEM);
643	}
644
645	return (0);
646}
647
648#ifndef __FreeBSD__
649void
650pfsync_state_export(struct pfsync_state *sp, struct pf_state *st)
651{
652	bzero(sp, sizeof(struct pfsync_state));
653
654	/* copy from state key */
655	sp->key[PF_SK_WIRE].addr[0] = st->key[PF_SK_WIRE]->addr[0];
656	sp->key[PF_SK_WIRE].addr[1] = st->key[PF_SK_WIRE]->addr[1];
657	sp->key[PF_SK_WIRE].port[0] = st->key[PF_SK_WIRE]->port[0];
658	sp->key[PF_SK_WIRE].port[1] = st->key[PF_SK_WIRE]->port[1];
659	sp->key[PF_SK_STACK].addr[0] = st->key[PF_SK_STACK]->addr[0];
660	sp->key[PF_SK_STACK].addr[1] = st->key[PF_SK_STACK]->addr[1];
661	sp->key[PF_SK_STACK].port[0] = st->key[PF_SK_STACK]->port[0];
662	sp->key[PF_SK_STACK].port[1] = st->key[PF_SK_STACK]->port[1];
663	sp->proto = st->key[PF_SK_WIRE]->proto;
664	sp->af = st->key[PF_SK_WIRE]->af;
665
666	/* copy from state */
667	strlcpy(sp->ifname, st->kif->pfik_name, sizeof(sp->ifname));
668	bcopy(&st->rt_addr, &sp->rt_addr, sizeof(sp->rt_addr));
669	sp->creation = htonl(time_second - st->creation);
670	sp->expire = pf_state_expires(st);
671	if (sp->expire <= time_second)
672		sp->expire = htonl(0);
673	else
674		sp->expire = htonl(sp->expire - time_second);
675
676	sp->direction = st->direction;
677	sp->log = st->log;
678	sp->timeout = st->timeout;
679	sp->state_flags = st->state_flags;
680	if (st->src_node)
681		sp->sync_flags |= PFSYNC_FLAG_SRCNODE;
682	if (st->nat_src_node)
683		sp->sync_flags |= PFSYNC_FLAG_NATSRCNODE;
684
685	bcopy(&st->id, &sp->id, sizeof(sp->id));
686	sp->creatorid = st->creatorid;
687	pf_state_peer_hton(&st->src, &sp->src);
688	pf_state_peer_hton(&st->dst, &sp->dst);
689
690	if (st->rule.ptr == NULL)
691		sp->rule = htonl(-1);
692	else
693		sp->rule = htonl(st->rule.ptr->nr);
694	if (st->anchor.ptr == NULL)
695		sp->anchor = htonl(-1);
696	else
697		sp->anchor = htonl(st->anchor.ptr->nr);
698	if (st->nat_rule.ptr == NULL)
699		sp->nat_rule = htonl(-1);
700	else
701		sp->nat_rule = htonl(st->nat_rule.ptr->nr);
702
703	pf_state_counter_hton(st->packets[0], sp->packets[0]);
704	pf_state_counter_hton(st->packets[1], sp->packets[1]);
705	pf_state_counter_hton(st->bytes[0], sp->bytes[0]);
706	pf_state_counter_hton(st->bytes[1], sp->bytes[1]);
707
708}
709#endif
710
711int
712pfsync_state_import(struct pfsync_state *sp, u_int8_t flags)
713{
714	struct pf_state	*st = NULL;
715	struct pf_state_key *skw = NULL, *sks = NULL;
716	struct pf_rule *r = NULL;
717	struct pfi_kif	*kif;
718	int pool_flags;
719	int error;
720
721	PF_LOCK_ASSERT();
722
723#ifdef __FreeBSD__
724	if (sp->creatorid == 0 && V_pf_status.debug >= PF_DEBUG_MISC) {
725#else
726	if (sp->creatorid == 0 && pf_status.debug >= PF_DEBUG_MISC) {
727#endif
728		printf("pfsync_state_import: invalid creator id:"
729		    " %08x\n", ntohl(sp->creatorid));
730		return (EINVAL);
731	}
732
733	if ((kif = pfi_kif_get(sp->ifname)) == NULL) {
734#ifdef __FreeBSD__
735		if (V_pf_status.debug >= PF_DEBUG_MISC)
736#else
737		if (pf_status.debug >= PF_DEBUG_MISC)
738#endif
739			printf("pfsync_state_import: "
740			    "unknown interface: %s\n", sp->ifname);
741		if (flags & PFSYNC_SI_IOCTL)
742			return (EINVAL);
743		return (0);	/* skip this state */
744	}
745
746	/*
747	 * If the ruleset checksums match or the state is coming from the ioctl,
748	 * it's safe to associate the state with the rule of that number.
749	 */
750	if (sp->rule != htonl(-1) && sp->anchor == htonl(-1) &&
751	    (flags & (PFSYNC_SI_IOCTL | PFSYNC_SI_CKSUM)) && ntohl(sp->rule) <
752	    pf_main_ruleset.rules[PF_RULESET_FILTER].active.rcount)
753		r = pf_main_ruleset.rules[
754		    PF_RULESET_FILTER].active.ptr_array[ntohl(sp->rule)];
755	else
756#ifdef __FreeBSD__
757		r = &V_pf_default_rule;
758#else
759		r = &pf_default_rule;
760#endif
761
762	if ((r->max_states && r->states_cur >= r->max_states))
763		goto cleanup;
764
765#ifdef __FreeBSD__
766	if (flags & PFSYNC_SI_IOCTL)
767		pool_flags = PR_WAITOK | PR_ZERO;
768	else
769		pool_flags = PR_NOWAIT | PR_ZERO;
770
771	if ((st = pool_get(&V_pf_state_pl, pool_flags)) == NULL)
772		goto cleanup;
773#else
774	if (flags & PFSYNC_SI_IOCTL)
775		pool_flags = PR_WAITOK | PR_LIMITFAIL | PR_ZERO;
776	else
777		pool_flags = PR_LIMITFAIL | PR_ZERO;
778
779	if ((st = pool_get(&pf_state_pl, pool_flags)) == NULL)
780		goto cleanup;
781#endif
782
783	if ((skw = pf_alloc_state_key(pool_flags)) == NULL)
784		goto cleanup;
785
786	if (PF_ANEQ(&sp->key[PF_SK_WIRE].addr[0],
787	    &sp->key[PF_SK_STACK].addr[0], sp->af) ||
788	    PF_ANEQ(&sp->key[PF_SK_WIRE].addr[1],
789	    &sp->key[PF_SK_STACK].addr[1], sp->af) ||
790	    sp->key[PF_SK_WIRE].port[0] != sp->key[PF_SK_STACK].port[0] ||
791	    sp->key[PF_SK_WIRE].port[1] != sp->key[PF_SK_STACK].port[1]) {
792		if ((sks = pf_alloc_state_key(pool_flags)) == NULL)
793			goto cleanup;
794	} else
795		sks = skw;
796
797	/* allocate memory for scrub info */
798	if (pfsync_alloc_scrub_memory(&sp->src, &st->src) ||
799	    pfsync_alloc_scrub_memory(&sp->dst, &st->dst))
800		goto cleanup;
801
802	/* copy to state key(s) */
803	skw->addr[0] = sp->key[PF_SK_WIRE].addr[0];
804	skw->addr[1] = sp->key[PF_SK_WIRE].addr[1];
805	skw->port[0] = sp->key[PF_SK_WIRE].port[0];
806	skw->port[1] = sp->key[PF_SK_WIRE].port[1];
807	skw->proto = sp->proto;
808	skw->af = sp->af;
809	if (sks != skw) {
810		sks->addr[0] = sp->key[PF_SK_STACK].addr[0];
811		sks->addr[1] = sp->key[PF_SK_STACK].addr[1];
812		sks->port[0] = sp->key[PF_SK_STACK].port[0];
813		sks->port[1] = sp->key[PF_SK_STACK].port[1];
814		sks->proto = sp->proto;
815		sks->af = sp->af;
816	}
817
818	/* copy to state */
819	bcopy(&sp->rt_addr, &st->rt_addr, sizeof(st->rt_addr));
820	st->creation = time_second - ntohl(sp->creation);
821	st->expire = time_second;
822	if (sp->expire) {
823		/* XXX No adaptive scaling. */
824		st->expire -= r->timeout[sp->timeout] - ntohl(sp->expire);
825	}
826
827	st->expire = ntohl(sp->expire) + time_second;
828	st->direction = sp->direction;
829	st->log = sp->log;
830	st->timeout = sp->timeout;
831	st->state_flags = sp->state_flags;
832
833	bcopy(sp->id, &st->id, sizeof(st->id));
834	st->creatorid = sp->creatorid;
835	pf_state_peer_ntoh(&sp->src, &st->src);
836	pf_state_peer_ntoh(&sp->dst, &st->dst);
837
838	st->rule.ptr = r;
839	st->nat_rule.ptr = NULL;
840	st->anchor.ptr = NULL;
841	st->rt_kif = NULL;
842
843	st->pfsync_time = time_second;
844	st->sync_state = PFSYNC_S_NONE;
845
846	/* XXX when we have nat_rule/anchors, use STATE_INC_COUNTERS */
847	r->states_cur++;
848	r->states_tot++;
849
850	if (!ISSET(flags, PFSYNC_SI_IOCTL))
851		SET(st->state_flags, PFSTATE_NOSYNC);
852
853	if ((error = pf_state_insert(kif, skw, sks, st)) != 0) {
854		/* XXX when we have nat_rule/anchors, use STATE_DEC_COUNTERS */
855		r->states_cur--;
856		goto cleanup_state;
857	}
858
859	if (!ISSET(flags, PFSYNC_SI_IOCTL)) {
860		CLR(st->state_flags, PFSTATE_NOSYNC);
861		if (ISSET(st->state_flags, PFSTATE_ACK)) {
862			pfsync_q_ins(st, PFSYNC_S_IACK);
863#ifdef __FreeBSD__
864			pfsync_sendout();
865#else
866			schednetisr(NETISR_PFSYNC);
867#endif
868		}
869	}
870	CLR(st->state_flags, PFSTATE_ACK);
871
872	return (0);
873
874cleanup:
875	error = ENOMEM;
876	if (skw == sks)
877		sks = NULL;
878#ifdef __FreeBSD__
879	if (skw != NULL)
880		pool_put(&V_pf_state_key_pl, skw);
881	if (sks != NULL)
882		pool_put(&V_pf_state_key_pl, sks);
883#else
884	if (skw != NULL)
885		pool_put(&pf_state_key_pl, skw);
886	if (sks != NULL)
887		pool_put(&pf_state_key_pl, sks);
888#endif
889
890cleanup_state:	/* pf_state_insert frees the state keys */
891	if (st) {
892#ifdef __FreeBSD__
893		if (st->dst.scrub)
894			pool_put(&V_pf_state_scrub_pl, st->dst.scrub);
895		if (st->src.scrub)
896			pool_put(&V_pf_state_scrub_pl, st->src.scrub);
897		pool_put(&V_pf_state_pl, st);
898#else
899		if (st->dst.scrub)
900			pool_put(&pf_state_scrub_pl, st->dst.scrub);
901		if (st->src.scrub)
902			pool_put(&pf_state_scrub_pl, st->src.scrub);
903		pool_put(&pf_state_pl, st);
904#endif
905	}
906	return (error);
907}
908
909void
910#ifdef __FreeBSD__
911pfsync_input(struct mbuf *m, __unused int off)
912#else
913pfsync_input(struct mbuf *m, ...)
914#endif
915{
916#ifdef __FreeBSD__
917	struct pfsync_softc *sc = V_pfsyncif;
918#else
919	struct pfsync_softc *sc = pfsyncif;
920#endif
921	struct pfsync_pkt pkt;
922	struct ip *ip = mtod(m, struct ip *);
923	struct pfsync_header *ph;
924	struct pfsync_subheader subh;
925
926	int offset;
927	int rv;
928
929	V_pfsyncstats.pfsyncs_ipackets++;
930
931	/* verify that we have a sync interface configured */
932#ifdef __FreeBSD__
933	if (!sc || !sc->sc_sync_if || !V_pf_status.running)
934#else
935	if (!sc || !sc->sc_sync_if || !pf_status.running)
936#endif
937		goto done;
938
939	/* verify that the packet came in on the right interface */
940	if (sc->sc_sync_if != m->m_pkthdr.rcvif) {
941		V_pfsyncstats.pfsyncs_badif++;
942		goto done;
943	}
944
945#ifdef __FreeBSD__
946	sc->sc_ifp->if_ipackets++;
947	sc->sc_ifp->if_ibytes += m->m_pkthdr.len;
948#else
949	sc->sc_if.if_ipackets++;
950	sc->sc_if.if_ibytes += m->m_pkthdr.len;
951#endif
952	/* verify that the IP TTL is 255. */
953	if (ip->ip_ttl != PFSYNC_DFLTTL) {
954		V_pfsyncstats.pfsyncs_badttl++;
955		goto done;
956	}
957
958	offset = ip->ip_hl << 2;
959	if (m->m_pkthdr.len < offset + sizeof(*ph)) {
960		V_pfsyncstats.pfsyncs_hdrops++;
961		goto done;
962	}
963
964	if (offset + sizeof(*ph) > m->m_len) {
965		if (m_pullup(m, offset + sizeof(*ph)) == NULL) {
966			V_pfsyncstats.pfsyncs_hdrops++;
967			return;
968		}
969		ip = mtod(m, struct ip *);
970	}
971	ph = (struct pfsync_header *)((char *)ip + offset);
972
973	/* verify the version */
974	if (ph->version != PFSYNC_VERSION) {
975		V_pfsyncstats.pfsyncs_badver++;
976		goto done;
977	}
978
979#if 0
980	if (pfsync_input_hmac(m, offset) != 0) {
981		/* XXX stats */
982		goto done;
983	}
984#endif
985
986	/* Cheaper to grab this now than having to mess with mbufs later */
987	pkt.ip = ip;
988	pkt.src = ip->ip_src;
989	pkt.flags = 0;
990
991#ifdef __FreeBSD__
992	if (!bcmp(&ph->pfcksum, &V_pf_status.pf_chksum, PF_MD5_DIGEST_LENGTH))
993#else
994	if (!bcmp(&ph->pfcksum, &pf_status.pf_chksum, PF_MD5_DIGEST_LENGTH))
995#endif
996		pkt.flags |= PFSYNC_SI_CKSUM;
997
998	offset += sizeof(*ph);
999	for (;;) {
1000		m_copydata(m, offset, sizeof(subh), (caddr_t)&subh);
1001		offset += sizeof(subh);
1002
1003		if (subh.action >= PFSYNC_ACT_MAX) {
1004			V_pfsyncstats.pfsyncs_badact++;
1005			goto done;
1006		}
1007
1008		rv = (*pfsync_acts[subh.action])(&pkt, m, offset,
1009		    ntohs(subh.count));
1010		if (rv == -1)
1011			return;
1012
1013		offset += rv;
1014	}
1015
1016done:
1017	m_freem(m);
1018}
1019
1020int
1021pfsync_in_clr(struct pfsync_pkt *pkt, struct mbuf *m, int offset, int count)
1022{
1023	struct pfsync_clr *clr;
1024	struct mbuf *mp;
1025	int len = sizeof(*clr) * count;
1026	int i, offp;
1027
1028	struct pf_state *st, *nexts;
1029	struct pf_state_key *sk, *nextsk;
1030	struct pf_state_item *si;
1031	u_int32_t creatorid;
1032	int s;
1033
1034	mp = m_pulldown(m, offset, len, &offp);
1035	if (mp == NULL) {
1036		V_pfsyncstats.pfsyncs_badlen++;
1037		return (-1);
1038	}
1039	clr = (struct pfsync_clr *)(mp->m_data + offp);
1040
1041	s = splsoftnet();
1042#ifdef __FreeBSD__
1043	PF_LOCK();
1044#endif
1045	for (i = 0; i < count; i++) {
1046		creatorid = clr[i].creatorid;
1047
1048		if (clr[i].ifname[0] == '\0') {
1049#ifdef __FreeBSD__
1050			for (st = RB_MIN(pf_state_tree_id, &V_tree_id);
1051			    st; st = nexts) {
1052				nexts = RB_NEXT(pf_state_tree_id, &V_tree_id, st);
1053#else
1054			for (st = RB_MIN(pf_state_tree_id, &tree_id);
1055			    st; st = nexts) {
1056				nexts = RB_NEXT(pf_state_tree_id, &tree_id, st);
1057#endif
1058				if (st->creatorid == creatorid) {
1059					SET(st->state_flags, PFSTATE_NOSYNC);
1060					pf_unlink_state(st);
1061				}
1062			}
1063		} else {
1064			if (pfi_kif_get(clr[i].ifname) == NULL)
1065				continue;
1066
1067			/* XXX correct? */
1068#ifdef __FreeBSD__
1069			for (sk = RB_MIN(pf_state_tree, &V_pf_statetbl);
1070#else
1071			for (sk = RB_MIN(pf_state_tree, &pf_statetbl);
1072#endif
1073			    sk; sk = nextsk) {
1074				nextsk = RB_NEXT(pf_state_tree,
1075#ifdef __FreeBSD__
1076				    &V_pf_statetbl, sk);
1077#else
1078				    &pf_statetbl, sk);
1079#endif
1080				TAILQ_FOREACH(si, &sk->states, entry) {
1081					if (si->s->creatorid == creatorid) {
1082						SET(si->s->state_flags,
1083						    PFSTATE_NOSYNC);
1084						pf_unlink_state(si->s);
1085					}
1086				}
1087			}
1088		}
1089	}
1090#ifdef __FreeBSD__
1091	PF_UNLOCK();
1092#endif
1093	splx(s);
1094
1095	return (len);
1096}
1097
1098int
1099pfsync_in_ins(struct pfsync_pkt *pkt, struct mbuf *m, int offset, int count)
1100{
1101	struct mbuf *mp;
1102	struct pfsync_state *sa, *sp;
1103	int len = sizeof(*sp) * count;
1104	int i, offp;
1105
1106	int s;
1107
1108	mp = m_pulldown(m, offset, len, &offp);
1109	if (mp == NULL) {
1110		V_pfsyncstats.pfsyncs_badlen++;
1111		return (-1);
1112	}
1113	sa = (struct pfsync_state *)(mp->m_data + offp);
1114
1115	s = splsoftnet();
1116#ifdef __FreeBSD__
1117	PF_LOCK();
1118#endif
1119	for (i = 0; i < count; i++) {
1120		sp = &sa[i];
1121
1122		/* check for invalid values */
1123		if (sp->timeout >= PFTM_MAX ||
1124		    sp->src.state > PF_TCPS_PROXY_DST ||
1125		    sp->dst.state > PF_TCPS_PROXY_DST ||
1126		    sp->direction > PF_OUT ||
1127		    (sp->af != AF_INET && sp->af != AF_INET6)) {
1128#ifdef __FreeBSD__
1129			if (V_pf_status.debug >= PF_DEBUG_MISC) {
1130#else
1131			if (pf_status.debug >= PF_DEBUG_MISC) {
1132#endif
1133				printf("pfsync_input: PFSYNC5_ACT_INS: "
1134				    "invalid value\n");
1135			}
1136			V_pfsyncstats.pfsyncs_badval++;
1137			continue;
1138		}
1139
1140		if (pfsync_state_import(sp, pkt->flags) == ENOMEM) {
1141			/* drop out, but process the rest of the actions */
1142			break;
1143		}
1144	}
1145#ifdef __FreeBSD__
1146	PF_UNLOCK();
1147#endif
1148	splx(s);
1149
1150	return (len);
1151}
1152
1153int
1154pfsync_in_iack(struct pfsync_pkt *pkt, struct mbuf *m, int offset, int count)
1155{
1156	struct pfsync_ins_ack *ia, *iaa;
1157	struct pf_state_cmp id_key;
1158	struct pf_state *st;
1159
1160	struct mbuf *mp;
1161	int len = count * sizeof(*ia);
1162	int offp, i;
1163	int s;
1164
1165	mp = m_pulldown(m, offset, len, &offp);
1166	if (mp == NULL) {
1167		V_pfsyncstats.pfsyncs_badlen++;
1168		return (-1);
1169	}
1170	iaa = (struct pfsync_ins_ack *)(mp->m_data + offp);
1171
1172	s = splsoftnet();
1173#ifdef __FreeBSD__
1174	PF_LOCK();
1175#endif
1176	for (i = 0; i < count; i++) {
1177		ia = &iaa[i];
1178
1179		bcopy(&ia->id, &id_key.id, sizeof(id_key.id));
1180		id_key.creatorid = ia->creatorid;
1181
1182		st = pf_find_state_byid(&id_key);
1183		if (st == NULL)
1184			continue;
1185
1186		if (ISSET(st->state_flags, PFSTATE_ACK))
1187			pfsync_deferred(st, 0);
1188	}
1189#ifdef __FreeBSD__
1190	PF_UNLOCK();
1191#endif
1192	splx(s);
1193	/*
1194	 * XXX this is not yet implemented, but we know the size of the
1195	 * message so we can skip it.
1196	 */
1197
1198	return (count * sizeof(struct pfsync_ins_ack));
1199}
1200
1201int
1202pfsync_upd_tcp(struct pf_state *st, struct pfsync_state_peer *src,
1203    struct pfsync_state_peer *dst)
1204{
1205	int sfail = 0;
1206
1207	/*
1208	 * The state should never go backwards except
1209	 * for syn-proxy states.  Neither should the
1210	 * sequence window slide backwards.
1211	 */
1212	if (st->src.state > src->state &&
1213	    (st->src.state < PF_TCPS_PROXY_SRC ||
1214	    src->state >= PF_TCPS_PROXY_SRC))
1215		sfail = 1;
1216	else if (SEQ_GT(st->src.seqlo, ntohl(src->seqlo)))
1217		sfail = 3;
1218	else if (st->dst.state > dst->state) {
1219		/* There might still be useful
1220		 * information about the src state here,
1221		 * so import that part of the update,
1222		 * then "fail" so we send the updated
1223		 * state back to the peer who is missing
1224		 * our what we know. */
1225		pf_state_peer_ntoh(src, &st->src);
1226		/* XXX do anything with timeouts? */
1227		sfail = 7;
1228	} else if (st->dst.state >= TCPS_SYN_SENT &&
1229	    SEQ_GT(st->dst.seqlo, ntohl(dst->seqlo)))
1230		sfail = 4;
1231
1232	return (sfail);
1233}
1234
1235int
1236pfsync_in_upd(struct pfsync_pkt *pkt, struct mbuf *m, int offset, int count)
1237{
1238	struct pfsync_state *sa, *sp;
1239	struct pf_state_cmp id_key;
1240	struct pf_state_key *sk;
1241	struct pf_state *st;
1242	int sfail;
1243
1244	struct mbuf *mp;
1245	int len = count * sizeof(*sp);
1246	int offp, i;
1247	int s;
1248
1249	mp = m_pulldown(m, offset, len, &offp);
1250	if (mp == NULL) {
1251		V_pfsyncstats.pfsyncs_badlen++;
1252		return (-1);
1253	}
1254	sa = (struct pfsync_state *)(mp->m_data + offp);
1255
1256	s = splsoftnet();
1257#ifdef __FreeBSD__
1258	PF_LOCK();
1259#endif
1260	for (i = 0; i < count; i++) {
1261		sp = &sa[i];
1262
1263		/* check for invalid values */
1264		if (sp->timeout >= PFTM_MAX ||
1265		    sp->src.state > PF_TCPS_PROXY_DST ||
1266		    sp->dst.state > PF_TCPS_PROXY_DST) {
1267#ifdef __FreeBSD__
1268			if (V_pf_status.debug >= PF_DEBUG_MISC) {
1269#else
1270			if (pf_status.debug >= PF_DEBUG_MISC) {
1271#endif
1272				printf("pfsync_input: PFSYNC_ACT_UPD: "
1273				    "invalid value\n");
1274			}
1275			V_pfsyncstats.pfsyncs_badval++;
1276			continue;
1277		}
1278
1279		bcopy(sp->id, &id_key.id, sizeof(id_key.id));
1280		id_key.creatorid = sp->creatorid;
1281
1282		st = pf_find_state_byid(&id_key);
1283		if (st == NULL) {
1284			/* insert the update */
1285			if (pfsync_state_import(sp, 0))
1286				V_pfsyncstats.pfsyncs_badstate++;
1287			continue;
1288		}
1289
1290		if (ISSET(st->state_flags, PFSTATE_ACK))
1291			pfsync_deferred(st, 1);
1292
1293		sk = st->key[PF_SK_WIRE];	/* XXX right one? */
1294		sfail = 0;
1295		if (sk->proto == IPPROTO_TCP)
1296			sfail = pfsync_upd_tcp(st, &sp->src, &sp->dst);
1297		else {
1298			/*
1299			 * Non-TCP protocol state machine always go
1300			 * forwards
1301			 */
1302			if (st->src.state > sp->src.state)
1303				sfail = 5;
1304			else if (st->dst.state > sp->dst.state)
1305				sfail = 6;
1306		}
1307
1308		if (sfail) {
1309#ifdef __FreeBSD__
1310			if (V_pf_status.debug >= PF_DEBUG_MISC) {
1311#else
1312			if (pf_status.debug >= PF_DEBUG_MISC) {
1313#endif
1314				printf("pfsync: %s stale update (%d)"
1315				    " id: %016llx creatorid: %08x\n",
1316				    (sfail < 7 ?  "ignoring" : "partial"),
1317				    sfail, betoh64(st->id),
1318				    ntohl(st->creatorid));
1319			}
1320			V_pfsyncstats.pfsyncs_stale++;
1321
1322			pfsync_update_state(st);
1323#ifdef __FreeBSD__
1324			pfsync_sendout();
1325#else
1326			schednetisr(NETISR_PFSYNC);
1327#endif
1328			continue;
1329		}
1330		pfsync_alloc_scrub_memory(&sp->dst, &st->dst);
1331		pf_state_peer_ntoh(&sp->src, &st->src);
1332		pf_state_peer_ntoh(&sp->dst, &st->dst);
1333		st->expire = ntohl(sp->expire) + time_second;
1334		st->timeout = sp->timeout;
1335		st->pfsync_time = time_second;
1336	}
1337#ifdef __FreeBSD__
1338	PF_UNLOCK();
1339#endif
1340	splx(s);
1341
1342	return (len);
1343}
1344
1345int
1346pfsync_in_upd_c(struct pfsync_pkt *pkt, struct mbuf *m, int offset, int count)
1347{
1348	struct pfsync_upd_c *ua, *up;
1349	struct pf_state_key *sk;
1350	struct pf_state_cmp id_key;
1351	struct pf_state *st;
1352
1353	int len = count * sizeof(*up);
1354	int sfail;
1355
1356	struct mbuf *mp;
1357	int offp, i;
1358	int s;
1359
1360	mp = m_pulldown(m, offset, len, &offp);
1361	if (mp == NULL) {
1362		V_pfsyncstats.pfsyncs_badlen++;
1363		return (-1);
1364	}
1365	ua = (struct pfsync_upd_c *)(mp->m_data + offp);
1366
1367	s = splsoftnet();
1368#ifdef __FreeBSD__
1369	PF_LOCK();
1370#endif
1371	for (i = 0; i < count; i++) {
1372		up = &ua[i];
1373
1374		/* check for invalid values */
1375		if (up->timeout >= PFTM_MAX ||
1376		    up->src.state > PF_TCPS_PROXY_DST ||
1377		    up->dst.state > PF_TCPS_PROXY_DST) {
1378#ifdef __FreeBSD__
1379			if (V_pf_status.debug >= PF_DEBUG_MISC) {
1380#else
1381			if (pf_status.debug >= PF_DEBUG_MISC) {
1382#endif
1383				printf("pfsync_input: "
1384				    "PFSYNC_ACT_UPD_C: "
1385				    "invalid value\n");
1386			}
1387			V_pfsyncstats.pfsyncs_badval++;
1388			continue;
1389		}
1390
1391		bcopy(&up->id, &id_key.id, sizeof(id_key.id));
1392		id_key.creatorid = up->creatorid;
1393
1394		st = pf_find_state_byid(&id_key);
1395		if (st == NULL) {
1396			/* We don't have this state. Ask for it. */
1397			pfsync_request_update(id_key.creatorid, id_key.id);
1398			continue;
1399		}
1400
1401		if (ISSET(st->state_flags, PFSTATE_ACK))
1402			pfsync_deferred(st, 1);
1403
1404		sk = st->key[PF_SK_WIRE]; /* XXX right one? */
1405		sfail = 0;
1406		if (sk->proto == IPPROTO_TCP)
1407			sfail = pfsync_upd_tcp(st, &up->src, &up->dst);
1408		else {
1409			/*
1410			 * Non-TCP protocol state machine always go forwards
1411			 */
1412			if (st->src.state > up->src.state)
1413				sfail = 5;
1414			else if (st->dst.state > up->dst.state)
1415				sfail = 6;
1416		}
1417
1418		if (sfail) {
1419#ifdef __FreeBSD__
1420			if (V_pf_status.debug >= PF_DEBUG_MISC) {
1421#else
1422			if (pf_status.debug >= PF_DEBUG_MISC) {
1423#endif
1424				printf("pfsync: ignoring stale update "
1425				    "(%d) id: %016llx "
1426				    "creatorid: %08x\n", sfail,
1427				    betoh64(st->id),
1428				    ntohl(st->creatorid));
1429			}
1430			V_pfsyncstats.pfsyncs_stale++;
1431
1432			pfsync_update_state(st);
1433#ifdef __FreeBSD__
1434			pfsync_sendout();
1435#else
1436			schednetisr(NETISR_PFSYNC);
1437#endif
1438			continue;
1439		}
1440		pfsync_alloc_scrub_memory(&up->dst, &st->dst);
1441		pf_state_peer_ntoh(&up->src, &st->src);
1442		pf_state_peer_ntoh(&up->dst, &st->dst);
1443		st->expire = ntohl(up->expire) + time_second;
1444		st->timeout = up->timeout;
1445		st->pfsync_time = time_second;
1446	}
1447#ifdef __FreeBSD__
1448	PF_UNLOCK();
1449#endif
1450	splx(s);
1451
1452	return (len);
1453}
1454
1455int
1456pfsync_in_ureq(struct pfsync_pkt *pkt, struct mbuf *m, int offset, int count)
1457{
1458	struct pfsync_upd_req *ur, *ura;
1459	struct mbuf *mp;
1460	int len = count * sizeof(*ur);
1461	int i, offp;
1462
1463	struct pf_state_cmp id_key;
1464	struct pf_state *st;
1465
1466	mp = m_pulldown(m, offset, len, &offp);
1467	if (mp == NULL) {
1468		V_pfsyncstats.pfsyncs_badlen++;
1469		return (-1);
1470	}
1471	ura = (struct pfsync_upd_req *)(mp->m_data + offp);
1472
1473	for (i = 0; i < count; i++) {
1474		ur = &ura[i];
1475
1476		bcopy(&ur->id, &id_key.id, sizeof(id_key.id));
1477		id_key.creatorid = ur->creatorid;
1478
1479		if (id_key.id == 0 && id_key.creatorid == 0)
1480			pfsync_bulk_start();
1481		else {
1482			st = pf_find_state_byid(&id_key);
1483			if (st == NULL) {
1484				V_pfsyncstats.pfsyncs_badstate++;
1485				continue;
1486			}
1487			if (ISSET(st->state_flags, PFSTATE_NOSYNC))
1488				continue;
1489
1490			PF_LOCK();
1491			pfsync_update_state_req(st);
1492			PF_UNLOCK();
1493		}
1494	}
1495
1496	return (len);
1497}
1498
1499int
1500pfsync_in_del(struct pfsync_pkt *pkt, struct mbuf *m, int offset, int count)
1501{
1502	struct mbuf *mp;
1503	struct pfsync_state *sa, *sp;
1504	struct pf_state_cmp id_key;
1505	struct pf_state *st;
1506	int len = count * sizeof(*sp);
1507	int offp, i;
1508	int s;
1509
1510	mp = m_pulldown(m, offset, len, &offp);
1511	if (mp == NULL) {
1512		V_pfsyncstats.pfsyncs_badlen++;
1513		return (-1);
1514	}
1515	sa = (struct pfsync_state *)(mp->m_data + offp);
1516
1517	s = splsoftnet();
1518#ifdef __FreeBSD__
1519	PF_LOCK();
1520#endif
1521	for (i = 0; i < count; i++) {
1522		sp = &sa[i];
1523
1524		bcopy(sp->id, &id_key.id, sizeof(id_key.id));
1525		id_key.creatorid = sp->creatorid;
1526
1527		st = pf_find_state_byid(&id_key);
1528		if (st == NULL) {
1529			V_pfsyncstats.pfsyncs_badstate++;
1530			continue;
1531		}
1532		SET(st->state_flags, PFSTATE_NOSYNC);
1533		pf_unlink_state(st);
1534	}
1535#ifdef __FreeBSD__
1536	PF_UNLOCK();
1537#endif
1538	splx(s);
1539
1540	return (len);
1541}
1542
1543int
1544pfsync_in_del_c(struct pfsync_pkt *pkt, struct mbuf *m, int offset, int count)
1545{
1546	struct mbuf *mp;
1547	struct pfsync_del_c *sa, *sp;
1548	struct pf_state_cmp id_key;
1549	struct pf_state *st;
1550	int len = count * sizeof(*sp);
1551	int offp, i;
1552	int s;
1553
1554	mp = m_pulldown(m, offset, len, &offp);
1555	if (mp == NULL) {
1556		V_pfsyncstats.pfsyncs_badlen++;
1557		return (-1);
1558	}
1559	sa = (struct pfsync_del_c *)(mp->m_data + offp);
1560
1561	s = splsoftnet();
1562#ifdef __FreeBSD__
1563	PF_LOCK();
1564#endif
1565	for (i = 0; i < count; i++) {
1566		sp = &sa[i];
1567
1568		bcopy(&sp->id, &id_key.id, sizeof(id_key.id));
1569		id_key.creatorid = sp->creatorid;
1570
1571		st = pf_find_state_byid(&id_key);
1572		if (st == NULL) {
1573			V_pfsyncstats.pfsyncs_badstate++;
1574			continue;
1575		}
1576
1577		SET(st->state_flags, PFSTATE_NOSYNC);
1578		pf_unlink_state(st);
1579	}
1580#ifdef __FreeBSD__
1581	PF_UNLOCK();
1582#endif
1583	splx(s);
1584
1585	return (len);
1586}
1587
1588int
1589pfsync_in_bus(struct pfsync_pkt *pkt, struct mbuf *m, int offset, int count)
1590{
1591#ifdef __FreeBSD__
1592	struct pfsync_softc *sc = V_pfsyncif;
1593#else
1594	struct pfsync_softc *sc = pfsyncif;
1595#endif
1596	struct pfsync_bus *bus;
1597	struct mbuf *mp;
1598	int len = count * sizeof(*bus);
1599	int offp;
1600
1601	/* If we're not waiting for a bulk update, who cares. */
1602	if (sc->sc_ureq_sent == 0)
1603		return (len);
1604
1605	mp = m_pulldown(m, offset, len, &offp);
1606	if (mp == NULL) {
1607		V_pfsyncstats.pfsyncs_badlen++;
1608		return (-1);
1609	}
1610	bus = (struct pfsync_bus *)(mp->m_data + offp);
1611
1612	switch (bus->status) {
1613	case PFSYNC_BUS_START:
1614#ifdef __FreeBSD__
1615		callout_reset(&sc->sc_bulkfail_tmo, 5 * hz, pfsync_bulk_fail,
1616		    V_pfsyncif);
1617#else
1618		timeout_add_sec(&sc->sc_bulkfail_tmo, 5); /* XXX magic */
1619#endif
1620#ifdef XXX
1621		    pf_pool_limits[PF_LIMIT_STATES].limit /
1622		    (PFSYNC_BULKPACKETS * sc->sc_maxcount));
1623#endif
1624#ifdef __FreeBSD__
1625		if (V_pf_status.debug >= PF_DEBUG_MISC)
1626#else
1627		if (pf_status.debug >= PF_DEBUG_MISC)
1628#endif
1629			printf("pfsync: received bulk update start\n");
1630		break;
1631
1632	case PFSYNC_BUS_END:
1633		if (time_uptime - ntohl(bus->endtime) >=
1634		    sc->sc_ureq_sent) {
1635			/* that's it, we're happy */
1636			sc->sc_ureq_sent = 0;
1637			sc->sc_bulk_tries = 0;
1638			timeout_del(&sc->sc_bulkfail_tmo);
1639#if NCARP > 0
1640#ifdef notyet
1641#ifdef __FreeBSD__
1642			if (!sc->pfsync_sync_ok)
1643#else
1644			if (!pfsync_sync_ok)
1645#endif
1646				carp_group_demote_adj(&sc->sc_if, -1);
1647#endif
1648#endif
1649#ifdef __FreeBSD__
1650			sc->pfsync_sync_ok = 1;
1651#else
1652			pfsync_sync_ok = 1;
1653#endif
1654#ifdef __FreeBSD__
1655			if (V_pf_status.debug >= PF_DEBUG_MISC)
1656#else
1657			if (pf_status.debug >= PF_DEBUG_MISC)
1658#endif
1659				printf("pfsync: received valid "
1660				    "bulk update end\n");
1661		} else {
1662#ifdef __FreeBSD__
1663			if (V_pf_status.debug >= PF_DEBUG_MISC)
1664#else
1665			if (pf_status.debug >= PF_DEBUG_MISC)
1666#endif
1667				printf("pfsync: received invalid "
1668				    "bulk update end: bad timestamp\n");
1669		}
1670		break;
1671	}
1672
1673	return (len);
1674}
1675
1676int
1677pfsync_in_tdb(struct pfsync_pkt *pkt, struct mbuf *m, int offset, int count)
1678{
1679	int len = count * sizeof(struct pfsync_tdb);
1680
1681#if defined(IPSEC)
1682	struct pfsync_tdb *tp;
1683	struct mbuf *mp;
1684	int offp;
1685	int i;
1686	int s;
1687
1688	mp = m_pulldown(m, offset, len, &offp);
1689	if (mp == NULL) {
1690		V_pfsyncstats.pfsyncs_badlen++;
1691		return (-1);
1692	}
1693	tp = (struct pfsync_tdb *)(mp->m_data + offp);
1694
1695	s = splsoftnet();
1696#ifdef __FreeBSD__
1697	PF_LOCK();
1698#endif
1699	for (i = 0; i < count; i++)
1700		pfsync_update_net_tdb(&tp[i]);
1701#ifdef __FreeBSD__
1702	PF_UNLOCK();
1703#endif
1704	splx(s);
1705#endif
1706
1707	return (len);
1708}
1709
1710#if defined(IPSEC)
1711/* Update an in-kernel tdb. Silently fail if no tdb is found. */
1712void
1713pfsync_update_net_tdb(struct pfsync_tdb *pt)
1714{
1715	struct tdb		*tdb;
1716	int			 s;
1717
1718	/* check for invalid values */
1719	if (ntohl(pt->spi) <= SPI_RESERVED_MAX ||
1720	    (pt->dst.sa.sa_family != AF_INET &&
1721	     pt->dst.sa.sa_family != AF_INET6))
1722		goto bad;
1723
1724	s = spltdb();
1725	tdb = gettdb(pt->spi, &pt->dst, pt->sproto);
1726	if (tdb) {
1727		pt->rpl = ntohl(pt->rpl);
1728		pt->cur_bytes = betoh64(pt->cur_bytes);
1729
1730		/* Neither replay nor byte counter should ever decrease. */
1731		if (pt->rpl < tdb->tdb_rpl ||
1732		    pt->cur_bytes < tdb->tdb_cur_bytes) {
1733			splx(s);
1734			goto bad;
1735		}
1736
1737		tdb->tdb_rpl = pt->rpl;
1738		tdb->tdb_cur_bytes = pt->cur_bytes;
1739	}
1740	splx(s);
1741	return;
1742
1743bad:
1744#ifdef __FreeBSD__
1745	if (V_pf_status.debug >= PF_DEBUG_MISC)
1746#else
1747	if (pf_status.debug >= PF_DEBUG_MISC)
1748#endif
1749		printf("pfsync_insert: PFSYNC_ACT_TDB_UPD: "
1750		    "invalid value\n");
1751	V_pfsyncstats.pfsyncs_badstate++;
1752	return;
1753}
1754#endif
1755
1756
1757int
1758pfsync_in_eof(struct pfsync_pkt *pkt, struct mbuf *m, int offset, int count)
1759{
1760	/* check if we are at the right place in the packet */
1761	if (offset != m->m_pkthdr.len - sizeof(struct pfsync_eof))
1762		V_pfsyncstats.pfsyncs_badact++;
1763
1764	/* we're done. free and let the caller return */
1765	m_freem(m);
1766	return (-1);
1767}
1768
1769int
1770pfsync_in_error(struct pfsync_pkt *pkt, struct mbuf *m, int offset, int count)
1771{
1772	V_pfsyncstats.pfsyncs_badact++;
1773
1774	m_freem(m);
1775	return (-1);
1776}
1777
1778int
1779pfsyncoutput(struct ifnet *ifp, struct mbuf *m, struct sockaddr *dst,
1780#ifdef __FreeBSD__
1781	struct route *rt)
1782#else
1783	struct rtentry *rt)
1784#endif
1785{
1786	m_freem(m);
1787	return (0);
1788}
1789
1790/* ARGSUSED */
1791int
1792pfsyncioctl(struct ifnet *ifp, u_long cmd, caddr_t data)
1793{
1794#ifndef __FreeBSD__
1795	struct proc *p = curproc;
1796#endif
1797	struct pfsync_softc *sc = ifp->if_softc;
1798	struct ifreq *ifr = (struct ifreq *)data;
1799	struct ip_moptions *imo = &sc->sc_imo;
1800	struct pfsyncreq pfsyncr;
1801	struct ifnet    *sifp;
1802	struct ip *ip;
1803	int s, error;
1804
1805	switch (cmd) {
1806#if 0
1807	case SIOCSIFADDR:
1808	case SIOCAIFADDR:
1809	case SIOCSIFDSTADDR:
1810#endif
1811	case SIOCSIFFLAGS:
1812#ifdef __FreeBSD__
1813		if (ifp->if_flags & IFF_UP)
1814			ifp->if_drv_flags |= IFF_DRV_RUNNING;
1815		else
1816			ifp->if_drv_flags &= ~IFF_DRV_RUNNING;
1817#else
1818		if (ifp->if_flags & IFF_UP)
1819			ifp->if_flags |= IFF_RUNNING;
1820		else
1821			ifp->if_flags &= ~IFF_RUNNING;
1822#endif
1823		break;
1824	case SIOCSIFMTU:
1825		if (ifr->ifr_mtu <= PFSYNC_MINPKT)
1826			return (EINVAL);
1827		if (ifr->ifr_mtu > MCLBYTES) /* XXX could be bigger */
1828			ifr->ifr_mtu = MCLBYTES;
1829		if (ifr->ifr_mtu < ifp->if_mtu) {
1830			s = splnet();
1831#ifdef __FreeBSD__
1832			PF_LOCK();
1833#endif
1834			pfsync_sendout();
1835#ifdef __FreeBSD__
1836			PF_UNLOCK();
1837#endif
1838			splx(s);
1839		}
1840		ifp->if_mtu = ifr->ifr_mtu;
1841		break;
1842	case SIOCGETPFSYNC:
1843		bzero(&pfsyncr, sizeof(pfsyncr));
1844		if (sc->sc_sync_if) {
1845			strlcpy(pfsyncr.pfsyncr_syncdev,
1846			    sc->sc_sync_if->if_xname, IFNAMSIZ);
1847		}
1848		pfsyncr.pfsyncr_syncpeer = sc->sc_sync_peer;
1849		pfsyncr.pfsyncr_maxupdates = sc->sc_maxupdates;
1850		return (copyout(&pfsyncr, ifr->ifr_data, sizeof(pfsyncr)));
1851
1852	case SIOCSETPFSYNC:
1853#ifdef __FreeBSD__
1854		if ((error = priv_check(curthread, PRIV_NETINET_PF)) != 0)
1855#else
1856		if ((error = suser(p, p->p_acflag)) != 0)
1857#endif
1858			return (error);
1859		if ((error = copyin(ifr->ifr_data, &pfsyncr, sizeof(pfsyncr))))
1860			return (error);
1861
1862#ifdef __FreeBSD__
1863		PF_LOCK();
1864#endif
1865		if (pfsyncr.pfsyncr_syncpeer.s_addr == 0)
1866#ifdef __FreeBSD__
1867			sc->sc_sync_peer.s_addr = htonl(INADDR_PFSYNC_GROUP);
1868#else
1869			sc->sc_sync_peer.s_addr = INADDR_PFSYNC_GROUP;
1870#endif
1871		else
1872			sc->sc_sync_peer.s_addr =
1873			    pfsyncr.pfsyncr_syncpeer.s_addr;
1874
1875		if (pfsyncr.pfsyncr_maxupdates > 255)
1876#ifdef __FreeBSD__
1877		{
1878			PF_UNLOCK();
1879#endif
1880			return (EINVAL);
1881#ifdef __FreeBSD__
1882		}
1883#endif
1884		sc->sc_maxupdates = pfsyncr.pfsyncr_maxupdates;
1885
1886		if (pfsyncr.pfsyncr_syncdev[0] == 0) {
1887			sc->sc_sync_if = NULL;
1888#ifdef __FreeBSD__
1889			PF_UNLOCK();
1890#endif
1891			if (imo->imo_num_memberships > 0) {
1892				in_delmulti(imo->imo_membership[
1893				    --imo->imo_num_memberships]);
1894				imo->imo_multicast_ifp = NULL;
1895			}
1896			break;
1897		}
1898
1899#ifdef __FreeBSD__
1900		PF_UNLOCK();
1901#endif
1902		if ((sifp = ifunit(pfsyncr.pfsyncr_syncdev)) == NULL)
1903			return (EINVAL);
1904
1905#ifdef __FreeBSD__
1906		PF_LOCK();
1907#endif
1908		s = splnet();
1909#ifdef __FreeBSD__
1910		if (sifp->if_mtu < sc->sc_ifp->if_mtu ||
1911#else
1912		if (sifp->if_mtu < sc->sc_if.if_mtu ||
1913#endif
1914		    (sc->sc_sync_if != NULL &&
1915		    sifp->if_mtu < sc->sc_sync_if->if_mtu) ||
1916		    sifp->if_mtu < MCLBYTES - sizeof(struct ip))
1917			pfsync_sendout();
1918		sc->sc_sync_if = sifp;
1919
1920		if (imo->imo_num_memberships > 0) {
1921#ifdef __FreeBSD__
1922			PF_UNLOCK();
1923#endif
1924			in_delmulti(imo->imo_membership[--imo->imo_num_memberships]);
1925#ifdef __FreeBSD__
1926			PF_LOCK();
1927#endif
1928			imo->imo_multicast_ifp = NULL;
1929		}
1930
1931		if (sc->sc_sync_if &&
1932#ifdef __FreeBSD__
1933		    sc->sc_sync_peer.s_addr == htonl(INADDR_PFSYNC_GROUP)) {
1934#else
1935		    sc->sc_sync_peer.s_addr == INADDR_PFSYNC_GROUP) {
1936#endif
1937			struct in_addr addr;
1938
1939			if (!(sc->sc_sync_if->if_flags & IFF_MULTICAST)) {
1940				sc->sc_sync_if = NULL;
1941#ifdef __FreeBSD__
1942				PF_UNLOCK();
1943#endif
1944				splx(s);
1945				return (EADDRNOTAVAIL);
1946			}
1947
1948#ifdef __FreeBSD__
1949			addr.s_addr = htonl(INADDR_PFSYNC_GROUP);
1950#else
1951			addr.s_addr = INADDR_PFSYNC_GROUP;
1952#endif
1953
1954#ifdef __FreeBSD__
1955			PF_UNLOCK();
1956#endif
1957			if ((imo->imo_membership[0] =
1958			    in_addmulti(&addr, sc->sc_sync_if)) == NULL) {
1959				sc->sc_sync_if = NULL;
1960				splx(s);
1961				return (ENOBUFS);
1962			}
1963#ifdef __FreeBSD__
1964			PF_LOCK();
1965#endif
1966			imo->imo_num_memberships++;
1967			imo->imo_multicast_ifp = sc->sc_sync_if;
1968			imo->imo_multicast_ttl = PFSYNC_DFLTTL;
1969			imo->imo_multicast_loop = 0;
1970		}
1971
1972		ip = &sc->sc_template;
1973		bzero(ip, sizeof(*ip));
1974		ip->ip_v = IPVERSION;
1975		ip->ip_hl = sizeof(sc->sc_template) >> 2;
1976		ip->ip_tos = IPTOS_LOWDELAY;
1977		/* len and id are set later */
1978#ifdef __FreeBSD__
1979		ip->ip_off = IP_DF;
1980#else
1981		ip->ip_off = htons(IP_DF);
1982#endif
1983		ip->ip_ttl = PFSYNC_DFLTTL;
1984		ip->ip_p = IPPROTO_PFSYNC;
1985		ip->ip_src.s_addr = INADDR_ANY;
1986		ip->ip_dst.s_addr = sc->sc_sync_peer.s_addr;
1987
1988		if (sc->sc_sync_if) {
1989			/* Request a full state table update. */
1990			sc->sc_ureq_sent = time_uptime;
1991#if NCARP > 0
1992#ifdef notyet
1993#ifdef __FreeBSD__
1994			if (sc->pfsync_sync_ok)
1995#else
1996			if (pfsync_sync_ok)
1997#endif
1998				carp_group_demote_adj(&sc->sc_if, 1);
1999#endif
2000#endif
2001#ifdef __FreeBSD__
2002			sc->pfsync_sync_ok = 0;
2003#else
2004			pfsync_sync_ok = 0;
2005#endif
2006#ifdef __FreeBSD__
2007			if (V_pf_status.debug >= PF_DEBUG_MISC)
2008#else
2009			if (pf_status.debug >= PF_DEBUG_MISC)
2010#endif
2011				printf("pfsync: requesting bulk update\n");
2012#ifdef __FreeBSD__
2013			callout_reset(&sc->sc_bulkfail_tmo, 5 * hz,
2014			    pfsync_bulk_fail, V_pfsyncif);
2015#else
2016			timeout_add_sec(&sc->sc_bulkfail_tmo, 5);
2017#endif
2018			pfsync_request_update(0, 0);
2019		}
2020#ifdef __FreeBSD__
2021		PF_UNLOCK();
2022#endif
2023		splx(s);
2024
2025		break;
2026
2027	default:
2028		return (ENOTTY);
2029	}
2030
2031	return (0);
2032}
2033
2034int
2035pfsync_out_state(struct pf_state *st, struct mbuf *m, int offset)
2036{
2037	struct pfsync_state *sp = (struct pfsync_state *)(m->m_data + offset);
2038
2039	pfsync_state_export(sp, st);
2040
2041	return (sizeof(*sp));
2042}
2043
2044int
2045pfsync_out_iack(struct pf_state *st, struct mbuf *m, int offset)
2046{
2047	struct pfsync_ins_ack *iack =
2048	    (struct pfsync_ins_ack *)(m->m_data + offset);
2049
2050	iack->id = st->id;
2051	iack->creatorid = st->creatorid;
2052
2053	return (sizeof(*iack));
2054}
2055
2056int
2057pfsync_out_upd_c(struct pf_state *st, struct mbuf *m, int offset)
2058{
2059	struct pfsync_upd_c *up = (struct pfsync_upd_c *)(m->m_data + offset);
2060
2061	up->id = st->id;
2062	pf_state_peer_hton(&st->src, &up->src);
2063	pf_state_peer_hton(&st->dst, &up->dst);
2064	up->creatorid = st->creatorid;
2065
2066	up->expire = pf_state_expires(st);
2067	if (up->expire <= time_second)
2068		up->expire = htonl(0);
2069	else
2070		up->expire = htonl(up->expire - time_second);
2071	up->timeout = st->timeout;
2072
2073	bzero(up->_pad, sizeof(up->_pad)); /* XXX */
2074
2075	return (sizeof(*up));
2076}
2077
2078int
2079pfsync_out_del(struct pf_state *st, struct mbuf *m, int offset)
2080{
2081	struct pfsync_del_c *dp = (struct pfsync_del_c *)(m->m_data + offset);
2082
2083	dp->id = st->id;
2084	dp->creatorid = st->creatorid;
2085
2086	SET(st->state_flags, PFSTATE_NOSYNC);
2087
2088	return (sizeof(*dp));
2089}
2090
2091void
2092pfsync_drop(struct pfsync_softc *sc)
2093{
2094	struct pf_state *st;
2095	struct pfsync_upd_req_item *ur;
2096#ifdef notyet
2097	struct tdb *t;
2098#endif
2099	int q;
2100
2101	for (q = 0; q < PFSYNC_S_COUNT; q++) {
2102		if (TAILQ_EMPTY(&sc->sc_qs[q]))
2103			continue;
2104
2105		TAILQ_FOREACH(st, &sc->sc_qs[q], sync_list) {
2106#ifdef PFSYNC_DEBUG
2107#ifdef __FreeBSD__
2108			KASSERT(st->sync_state == q,
2109				("%s: st->sync_state == q",
2110					__FUNCTION__));
2111#else
2112			KASSERT(st->sync_state == q);
2113#endif
2114#endif
2115			st->sync_state = PFSYNC_S_NONE;
2116		}
2117		TAILQ_INIT(&sc->sc_qs[q]);
2118	}
2119
2120	while ((ur = TAILQ_FIRST(&sc->sc_upd_req_list)) != NULL) {
2121		TAILQ_REMOVE(&sc->sc_upd_req_list, ur, ur_entry);
2122		pool_put(&sc->sc_pool, ur);
2123	}
2124
2125	sc->sc_plus = NULL;
2126
2127#ifdef notyet
2128	if (!TAILQ_EMPTY(&sc->sc_tdb_q)) {
2129		TAILQ_FOREACH(t, &sc->sc_tdb_q, tdb_sync_entry)
2130			CLR(t->tdb_flags, TDBF_PFSYNC);
2131
2132		TAILQ_INIT(&sc->sc_tdb_q);
2133	}
2134#endif
2135
2136	sc->sc_len = PFSYNC_MINPKT;
2137}
2138
2139void
2140pfsync_sendout(void)
2141{
2142#ifdef __FreeBSD__
2143	struct pfsync_softc *sc = V_pfsyncif;
2144#else
2145	struct pfsync_softc *sc = pfsyncif;
2146#endif
2147#if NBPFILTER > 0
2148#ifdef __FreeBSD__
2149	struct ifnet *ifp = sc->sc_ifp;
2150#else
2151	struct ifnet *ifp = &sc->sc_if;
2152#endif
2153#endif
2154	struct mbuf *m;
2155	struct ip *ip;
2156	struct pfsync_header *ph;
2157	struct pfsync_subheader *subh;
2158	struct pf_state *st;
2159	struct pfsync_upd_req_item *ur;
2160#ifdef notyet
2161	struct tdb *t;
2162#endif
2163#ifdef __FreeBSD__
2164	size_t pktlen;
2165	int dummy_error;
2166#endif
2167	int offset;
2168	int q, count = 0;
2169
2170#ifdef __FreeBSD__
2171	PF_LOCK_ASSERT();
2172#else
2173	splassert(IPL_NET);
2174#endif
2175
2176	if (sc == NULL || sc->sc_len == PFSYNC_MINPKT)
2177		return;
2178
2179#if NBPFILTER > 0
2180	if (ifp->if_bpf == NULL && sc->sc_sync_if == NULL) {
2181#else
2182	if (sc->sc_sync_if == NULL) {
2183#endif
2184		pfsync_drop(sc);
2185		return;
2186	}
2187
2188	MGETHDR(m, M_DONTWAIT, MT_DATA);
2189	if (m == NULL) {
2190#ifdef __FreeBSD__
2191		sc->sc_ifp->if_oerrors++;
2192#else
2193		sc->sc_if.if_oerrors++;
2194#endif
2195		V_pfsyncstats.pfsyncs_onomem++;
2196		pfsync_drop(sc);
2197		return;
2198	}
2199
2200#ifdef __FreeBSD__
2201	pktlen = max_linkhdr + sc->sc_len;
2202	if (pktlen > MHLEN) {
2203		/* Find the right pool to allocate from. */
2204		/* XXX: This is ugly. */
2205		m_cljget(m, M_DONTWAIT, pktlen <= MSIZE ? MSIZE :
2206			pktlen <= MCLBYTES ? MCLBYTES :
2207#if MJUMPAGESIZE != MCLBYTES
2208			pktlen <= MJUMPAGESIZE ? MJUMPAGESIZE :
2209#endif
2210			pktlen <= MJUM9BYTES ? MJUM9BYTES : MJUM16BYTES);
2211#else
2212	if (max_linkhdr + sc->sc_len > MHLEN) {
2213		MCLGETI(m, M_DONTWAIT, NULL, max_linkhdr + sc->sc_len);
2214#endif
2215		if (!ISSET(m->m_flags, M_EXT)) {
2216			m_free(m);
2217#ifdef __FreeBSD__
2218			sc->sc_ifp->if_oerrors++;
2219#else
2220			sc->sc_if.if_oerrors++;
2221#endif
2222			V_pfsyncstats.pfsyncs_onomem++;
2223			pfsync_drop(sc);
2224			return;
2225		}
2226	}
2227	m->m_data += max_linkhdr;
2228	m->m_len = m->m_pkthdr.len = sc->sc_len;
2229
2230	/* build the ip header */
2231	ip = (struct ip *)m->m_data;
2232	bcopy(&sc->sc_template, ip, sizeof(*ip));
2233	offset = sizeof(*ip);
2234
2235#ifdef __FreeBSD__
2236	ip->ip_len = m->m_pkthdr.len;
2237#else
2238	ip->ip_len = htons(m->m_pkthdr.len);
2239#endif
2240	ip->ip_id = htons(ip_randomid());
2241
2242	/* build the pfsync header */
2243	ph = (struct pfsync_header *)(m->m_data + offset);
2244	bzero(ph, sizeof(*ph));
2245	offset += sizeof(*ph);
2246
2247	ph->version = PFSYNC_VERSION;
2248	ph->len = htons(sc->sc_len - sizeof(*ip));
2249#ifdef __FreeBSD__
2250	bcopy(V_pf_status.pf_chksum, ph->pfcksum, PF_MD5_DIGEST_LENGTH);
2251#else
2252	bcopy(pf_status.pf_chksum, ph->pfcksum, PF_MD5_DIGEST_LENGTH);
2253#endif
2254
2255	/* walk the queues */
2256	for (q = 0; q < PFSYNC_S_COUNT; q++) {
2257		if (TAILQ_EMPTY(&sc->sc_qs[q]))
2258			continue;
2259
2260		subh = (struct pfsync_subheader *)(m->m_data + offset);
2261		offset += sizeof(*subh);
2262
2263		count = 0;
2264		TAILQ_FOREACH(st, &sc->sc_qs[q], sync_list) {
2265#ifdef PFSYNC_DEBUG
2266#ifdef __FreeBSD__
2267			KASSERT(st->sync_state == q,
2268				("%s: st->sync_state == q",
2269					__FUNCTION__));
2270#else
2271			KASSERT(st->sync_state == q);
2272#endif
2273#endif
2274
2275			offset += pfsync_qs[q].write(st, m, offset);
2276			st->sync_state = PFSYNC_S_NONE;
2277			count++;
2278		}
2279		TAILQ_INIT(&sc->sc_qs[q]);
2280
2281		bzero(subh, sizeof(*subh));
2282		subh->action = pfsync_qs[q].action;
2283		subh->count = htons(count);
2284	}
2285
2286	if (!TAILQ_EMPTY(&sc->sc_upd_req_list)) {
2287		subh = (struct pfsync_subheader *)(m->m_data + offset);
2288		offset += sizeof(*subh);
2289
2290		count = 0;
2291		while ((ur = TAILQ_FIRST(&sc->sc_upd_req_list)) != NULL) {
2292			TAILQ_REMOVE(&sc->sc_upd_req_list, ur, ur_entry);
2293
2294			bcopy(&ur->ur_msg, m->m_data + offset,
2295			    sizeof(ur->ur_msg));
2296			offset += sizeof(ur->ur_msg);
2297
2298			pool_put(&sc->sc_pool, ur);
2299
2300			count++;
2301		}
2302
2303		bzero(subh, sizeof(*subh));
2304		subh->action = PFSYNC_ACT_UPD_REQ;
2305		subh->count = htons(count);
2306	}
2307
2308	/* has someone built a custom region for us to add? */
2309	if (sc->sc_plus != NULL) {
2310		bcopy(sc->sc_plus, m->m_data + offset, sc->sc_pluslen);
2311		offset += sc->sc_pluslen;
2312
2313		sc->sc_plus = NULL;
2314	}
2315
2316#ifdef notyet
2317	if (!TAILQ_EMPTY(&sc->sc_tdb_q)) {
2318		subh = (struct pfsync_subheader *)(m->m_data + offset);
2319		offset += sizeof(*subh);
2320
2321		count = 0;
2322		TAILQ_FOREACH(t, &sc->sc_tdb_q, tdb_sync_entry) {
2323			offset += pfsync_out_tdb(t, m, offset);
2324			CLR(t->tdb_flags, TDBF_PFSYNC);
2325
2326			count++;
2327		}
2328		TAILQ_INIT(&sc->sc_tdb_q);
2329
2330		bzero(subh, sizeof(*subh));
2331		subh->action = PFSYNC_ACT_TDB;
2332		subh->count = htons(count);
2333	}
2334#endif
2335
2336	subh = (struct pfsync_subheader *)(m->m_data + offset);
2337	offset += sizeof(*subh);
2338
2339	bzero(subh, sizeof(*subh));
2340	subh->action = PFSYNC_ACT_EOF;
2341	subh->count = htons(1);
2342
2343	/* XXX write checksum in EOF here */
2344
2345	/* we're done, let's put it on the wire */
2346#if NBPFILTER > 0
2347	if (ifp->if_bpf) {
2348		m->m_data += sizeof(*ip);
2349		m->m_len = m->m_pkthdr.len = sc->sc_len - sizeof(*ip);
2350#ifdef __FreeBSD__
2351		BPF_MTAP(ifp, m);
2352#else
2353		bpf_mtap(ifp->if_bpf, m, BPF_DIRECTION_OUT);
2354#endif
2355		m->m_data -= sizeof(*ip);
2356		m->m_len = m->m_pkthdr.len = sc->sc_len;
2357	}
2358
2359	if (sc->sc_sync_if == NULL) {
2360		sc->sc_len = PFSYNC_MINPKT;
2361		m_freem(m);
2362		return;
2363	}
2364#endif
2365
2366#ifdef __FreeBSD__
2367	sc->sc_ifp->if_opackets++;
2368	sc->sc_ifp->if_obytes += m->m_pkthdr.len;
2369	sc->sc_len = PFSYNC_MINPKT;
2370
2371	IFQ_ENQUEUE(&sc->sc_ifp->if_snd, m, dummy_error);
2372	schednetisr(NETISR_PFSYNC);
2373#else
2374	sc->sc_if.if_opackets++;
2375	sc->sc_if.if_obytes += m->m_pkthdr.len;
2376
2377	if (ip_output(m, NULL, NULL, IP_RAWOUTPUT, &sc->sc_imo, NULL) == 0)
2378		pfsyncstats.pfsyncs_opackets++;
2379	else
2380		pfsyncstats.pfsyncs_oerrors++;
2381
2382	/* start again */
2383	sc->sc_len = PFSYNC_MINPKT;
2384#endif
2385}
2386
2387void
2388pfsync_insert_state(struct pf_state *st)
2389{
2390#ifdef __FreeBSD__
2391	struct pfsync_softc *sc = V_pfsyncif;
2392#else
2393	struct pfsync_softc *sc = pfsyncif;
2394#endif
2395
2396#ifdef __FreeBSD__
2397	PF_LOCK_ASSERT();
2398#else
2399	splassert(IPL_SOFTNET);
2400#endif
2401
2402	if (ISSET(st->rule.ptr->rule_flag, PFRULE_NOSYNC) ||
2403	    st->key[PF_SK_WIRE]->proto == IPPROTO_PFSYNC) {
2404		SET(st->state_flags, PFSTATE_NOSYNC);
2405		return;
2406	}
2407
2408	if (sc == NULL || ISSET(st->state_flags, PFSTATE_NOSYNC))
2409		return;
2410
2411#ifdef PFSYNC_DEBUG
2412#ifdef __FreeBSD__
2413	KASSERT(st->sync_state == PFSYNC_S_NONE,
2414		("%s: st->sync_state == PFSYNC_S_NONE", __FUNCTION__));
2415#else
2416	KASSERT(st->sync_state == PFSYNC_S_NONE);
2417#endif
2418#endif
2419
2420	if (sc->sc_len == PFSYNC_MINPKT)
2421#ifdef __FreeBSD__
2422		callout_reset(&sc->sc_tmo, 1 * hz, pfsync_timeout,
2423		    V_pfsyncif);
2424#else
2425		timeout_add_sec(&sc->sc_tmo, 1);
2426#endif
2427
2428	pfsync_q_ins(st, PFSYNC_S_INS);
2429
2430	if (ISSET(st->state_flags, PFSTATE_ACK))
2431#ifdef __FreeBSD__
2432		pfsync_sendout();
2433#else
2434		schednetisr(NETISR_PFSYNC);
2435#endif
2436	else
2437		st->sync_updates = 0;
2438}
2439
2440int defer = 10;
2441
2442int
2443pfsync_defer(struct pf_state *st, struct mbuf *m)
2444{
2445#ifdef __FreeBSD__
2446	struct pfsync_softc *sc = V_pfsyncif;
2447#else
2448	struct pfsync_softc *sc = pfsyncif;
2449#endif
2450	struct pfsync_deferral *pd;
2451
2452#ifdef __FreeBSD__
2453	PF_LOCK_ASSERT();
2454#else
2455	splassert(IPL_SOFTNET);
2456#endif
2457
2458	if (sc->sc_deferred >= 128)
2459		pfsync_undefer(TAILQ_FIRST(&sc->sc_deferrals), 0);
2460
2461	pd = pool_get(&sc->sc_pool, M_NOWAIT);
2462	if (pd == NULL)
2463		return (0);
2464	sc->sc_deferred++;
2465
2466#ifdef __FreeBSD__
2467	m->m_flags |= M_SKIP_FIREWALL;
2468#else
2469	m->m_pkthdr.pf.flags |= PF_TAG_GENERATED;
2470#endif
2471	SET(st->state_flags, PFSTATE_ACK);
2472
2473	pd->pd_st = st;
2474	pd->pd_m = m;
2475
2476	TAILQ_INSERT_TAIL(&sc->sc_deferrals, pd, pd_entry);
2477#ifdef __FreeBSD__
2478	callout_init(&pd->pd_tmo, CALLOUT_MPSAFE);
2479	callout_reset(&pd->pd_tmo, defer, pfsync_defer_tmo,
2480		pd);
2481#else
2482	timeout_set(&pd->pd_tmo, pfsync_defer_tmo, pd);
2483	timeout_add(&pd->pd_tmo, defer);
2484#endif
2485
2486	return (1);
2487}
2488
2489void
2490pfsync_undefer(struct pfsync_deferral *pd, int drop)
2491{
2492#ifdef __FreeBSD__
2493	struct pfsync_softc *sc = V_pfsyncif;
2494#else
2495	struct pfsync_softc *sc = pfsyncif;
2496#endif
2497	int s;
2498
2499#ifdef __FreeBSD__
2500	PF_LOCK_ASSERT();
2501#else
2502	splassert(IPL_SOFTNET);
2503#endif
2504
2505	TAILQ_REMOVE(&sc->sc_deferrals, pd, pd_entry);
2506	sc->sc_deferred--;
2507
2508	CLR(pd->pd_st->state_flags, PFSTATE_ACK);
2509	timeout_del(&pd->pd_tmo); /* bah */
2510	if (drop)
2511		m_freem(pd->pd_m);
2512	else {
2513		s = splnet();
2514#ifdef __FreeBSD__
2515		/* XXX: use pf_defered?! */
2516		PF_UNLOCK();
2517#endif
2518		ip_output(pd->pd_m, (void *)NULL, (void *)NULL, 0,
2519		    (void *)NULL, (void *)NULL);
2520#ifdef __FreeBSD__
2521		PF_LOCK();
2522#endif
2523		splx(s);
2524	}
2525
2526	pool_put(&sc->sc_pool, pd);
2527}
2528
2529void
2530pfsync_defer_tmo(void *arg)
2531{
2532#if defined(__FreeBSD__) && defined(VIMAGE)
2533	struct pfsync_deferral *pd = arg;
2534#endif
2535	int s;
2536
2537	s = splsoftnet();
2538#ifdef __FreeBSD__
2539	CURVNET_SET(pd->pd_m->m_pkthdr.rcvif->if_vnet); /* XXX */
2540	PF_LOCK();
2541#endif
2542	pfsync_undefer(arg, 0);
2543#ifdef __FreeBSD__
2544	PF_UNLOCK();
2545	CURVNET_RESTORE();
2546#endif
2547	splx(s);
2548}
2549
2550void
2551pfsync_deferred(struct pf_state *st, int drop)
2552{
2553#ifdef __FreeBSD__
2554	struct pfsync_softc *sc = V_pfsyncif;
2555#else
2556	struct pfsync_softc *sc = pfsyncif;
2557#endif
2558	struct pfsync_deferral *pd;
2559
2560	TAILQ_FOREACH(pd, &sc->sc_deferrals, pd_entry) {
2561		 if (pd->pd_st == st) {
2562			pfsync_undefer(pd, drop);
2563			return;
2564		}
2565	}
2566
2567	panic("pfsync_send_deferred: unable to find deferred state");
2568}
2569
2570u_int pfsync_upds = 0;
2571
2572void
2573pfsync_update_state(struct pf_state *st)
2574{
2575#ifdef __FreeBSD__
2576	struct pfsync_softc *sc = V_pfsyncif;
2577#else
2578	struct pfsync_softc *sc = pfsyncif;
2579#endif
2580	int sync = 0;
2581
2582#ifdef __FreeBSD__
2583	PF_LOCK_ASSERT();
2584#else
2585	splassert(IPL_SOFTNET);
2586#endif
2587
2588	if (sc == NULL)
2589		return;
2590
2591	if (ISSET(st->state_flags, PFSTATE_ACK))
2592		pfsync_deferred(st, 0);
2593	if (ISSET(st->state_flags, PFSTATE_NOSYNC)) {
2594		if (st->sync_state != PFSYNC_S_NONE)
2595			pfsync_q_del(st);
2596		return;
2597	}
2598
2599	if (sc->sc_len == PFSYNC_MINPKT)
2600#ifdef __FreeBSD__
2601		callout_reset(&sc->sc_tmo, 1 * hz, pfsync_timeout,
2602		    V_pfsyncif);
2603#else
2604		timeout_add_sec(&sc->sc_tmo, 1);
2605#endif
2606
2607	switch (st->sync_state) {
2608	case PFSYNC_S_UPD_C:
2609	case PFSYNC_S_UPD:
2610	case PFSYNC_S_INS:
2611		/* we're already handling it */
2612
2613		st->sync_updates++;
2614		if (st->sync_updates >= sc->sc_maxupdates)
2615			sync = 1;
2616		break;
2617
2618	case PFSYNC_S_IACK:
2619		pfsync_q_del(st);
2620	case PFSYNC_S_NONE:
2621		pfsync_q_ins(st, PFSYNC_S_UPD_C);
2622		st->sync_updates = 0;
2623		break;
2624
2625	default:
2626		panic("pfsync_update_state: unexpected sync state %d",
2627		    st->sync_state);
2628	}
2629
2630	if (sync || (time_second - st->pfsync_time) < 2) {
2631		pfsync_upds++;
2632#ifdef __FreeBSD__
2633		pfsync_sendout();
2634#else
2635		schednetisr(NETISR_PFSYNC);
2636#endif
2637	}
2638}
2639
2640void
2641pfsync_request_update(u_int32_t creatorid, u_int64_t id)
2642{
2643#ifdef __FreeBSD__
2644	struct pfsync_softc *sc = V_pfsyncif;
2645#else
2646	struct pfsync_softc *sc = pfsyncif;
2647#endif
2648	struct pfsync_upd_req_item *item;
2649	size_t nlen = sizeof(struct pfsync_upd_req);
2650	int s;
2651
2652	PF_LOCK_ASSERT();
2653
2654	/*
2655	 * this code does nothing to prevent multiple update requests for the
2656	 * same state being generated.
2657	 */
2658
2659	item = pool_get(&sc->sc_pool, PR_NOWAIT);
2660	if (item == NULL) {
2661		/* XXX stats */
2662		return;
2663	}
2664
2665	item->ur_msg.id = id;
2666	item->ur_msg.creatorid = creatorid;
2667
2668	if (TAILQ_EMPTY(&sc->sc_upd_req_list))
2669		nlen += sizeof(struct pfsync_subheader);
2670
2671#ifdef __FreeBSD__
2672	if (sc->sc_len + nlen > sc->sc_ifp->if_mtu) {
2673#else
2674	if (sc->sc_len + nlen > sc->sc_if.if_mtu) {
2675#endif
2676		s = splnet();
2677		pfsync_sendout();
2678		splx(s);
2679
2680		nlen = sizeof(struct pfsync_subheader) +
2681		    sizeof(struct pfsync_upd_req);
2682	}
2683
2684	TAILQ_INSERT_TAIL(&sc->sc_upd_req_list, item, ur_entry);
2685	sc->sc_len += nlen;
2686
2687#ifdef __FreeBSD__
2688	pfsync_sendout();
2689#else
2690	schednetisr(NETISR_PFSYNC);
2691#endif
2692}
2693
2694void
2695pfsync_update_state_req(struct pf_state *st)
2696{
2697#ifdef __FreeBSD__
2698	struct pfsync_softc *sc = V_pfsyncif;
2699#else
2700	struct pfsync_softc *sc = pfsyncif;
2701#endif
2702
2703	PF_LOCK_ASSERT();
2704
2705	if (sc == NULL)
2706		panic("pfsync_update_state_req: nonexistant instance");
2707
2708	if (ISSET(st->state_flags, PFSTATE_NOSYNC)) {
2709		if (st->sync_state != PFSYNC_S_NONE)
2710			pfsync_q_del(st);
2711		return;
2712	}
2713
2714	switch (st->sync_state) {
2715	case PFSYNC_S_UPD_C:
2716	case PFSYNC_S_IACK:
2717		pfsync_q_del(st);
2718	case PFSYNC_S_NONE:
2719		pfsync_q_ins(st, PFSYNC_S_UPD);
2720#ifdef __FreeBSD__
2721		pfsync_sendout();
2722#else
2723		schednetisr(NETISR_PFSYNC);
2724#endif
2725		return;
2726
2727	case PFSYNC_S_INS:
2728	case PFSYNC_S_UPD:
2729	case PFSYNC_S_DEL:
2730		/* we're already handling it */
2731		return;
2732
2733	default:
2734		panic("pfsync_update_state_req: unexpected sync state %d",
2735		    st->sync_state);
2736	}
2737}
2738
2739void
2740pfsync_delete_state(struct pf_state *st)
2741{
2742#ifdef __FreeBSD__
2743	struct pfsync_softc *sc = V_pfsyncif;
2744#else
2745	struct pfsync_softc *sc = pfsyncif;
2746#endif
2747
2748#ifdef __FreeBSD__
2749	PF_LOCK_ASSERT();
2750#else
2751	splassert(IPL_SOFTNET);
2752#endif
2753
2754	if (sc == NULL)
2755		return;
2756
2757	if (ISSET(st->state_flags, PFSTATE_ACK))
2758		pfsync_deferred(st, 1);
2759	if (ISSET(st->state_flags, PFSTATE_NOSYNC)) {
2760		if (st->sync_state != PFSYNC_S_NONE)
2761			pfsync_q_del(st);
2762		return;
2763	}
2764
2765	if (sc->sc_len == PFSYNC_MINPKT)
2766#ifdef __FreeBSD__
2767		callout_reset(&sc->sc_tmo, 1 * hz, pfsync_timeout,
2768		    V_pfsyncif);
2769#else
2770		timeout_add_sec(&sc->sc_tmo, 1);
2771#endif
2772
2773	switch (st->sync_state) {
2774	case PFSYNC_S_INS:
2775		/* we never got to tell the world so just forget about it */
2776		pfsync_q_del(st);
2777		return;
2778
2779	case PFSYNC_S_UPD_C:
2780	case PFSYNC_S_UPD:
2781	case PFSYNC_S_IACK:
2782		pfsync_q_del(st);
2783		/* FALLTHROUGH to putting it on the del list */
2784
2785	case PFSYNC_S_NONE:
2786		pfsync_q_ins(st, PFSYNC_S_DEL);
2787		return;
2788
2789	default:
2790		panic("pfsync_delete_state: unexpected sync state %d",
2791		    st->sync_state);
2792	}
2793}
2794
2795void
2796pfsync_clear_states(u_int32_t creatorid, const char *ifname)
2797{
2798	struct {
2799		struct pfsync_subheader subh;
2800		struct pfsync_clr clr;
2801	} __packed r;
2802
2803#ifdef __FreeBSD__
2804	struct pfsync_softc *sc = V_pfsyncif;
2805#else
2806	struct pfsync_softc *sc = pfsyncif;
2807#endif
2808
2809#ifdef __FreeBSD__
2810	PF_LOCK_ASSERT();
2811#else
2812	splassert(IPL_SOFTNET);
2813#endif
2814
2815	if (sc == NULL)
2816		return;
2817
2818	bzero(&r, sizeof(r));
2819
2820	r.subh.action = PFSYNC_ACT_CLR;
2821	r.subh.count = htons(1);
2822
2823	strlcpy(r.clr.ifname, ifname, sizeof(r.clr.ifname));
2824	r.clr.creatorid = creatorid;
2825
2826	pfsync_send_plus(&r, sizeof(r));
2827}
2828
2829void
2830pfsync_q_ins(struct pf_state *st, int q)
2831{
2832#ifdef __FreeBSD__
2833	struct pfsync_softc *sc = V_pfsyncif;
2834#else
2835	struct pfsync_softc *sc = pfsyncif;
2836#endif
2837	size_t nlen = pfsync_qs[q].len;
2838	int s;
2839
2840	PF_LOCK_ASSERT();
2841
2842#ifdef __FreeBSD__
2843	KASSERT(st->sync_state == PFSYNC_S_NONE,
2844		("%s: st->sync_state == PFSYNC_S_NONE", __FUNCTION__));
2845#else
2846	KASSERT(st->sync_state == PFSYNC_S_NONE);
2847#endif
2848
2849#if 1 || defined(PFSYNC_DEBUG)
2850	if (sc->sc_len < PFSYNC_MINPKT)
2851#ifdef __FreeBSD__
2852		panic("pfsync pkt len is too low %zu", sc->sc_len);
2853#else
2854		panic("pfsync pkt len is too low %d", sc->sc_len);
2855#endif
2856#endif
2857	if (TAILQ_EMPTY(&sc->sc_qs[q]))
2858		nlen += sizeof(struct pfsync_subheader);
2859
2860#ifdef __FreeBSD__
2861	if (sc->sc_len + nlen > sc->sc_ifp->if_mtu) {
2862#else
2863	if (sc->sc_len + nlen > sc->sc_if.if_mtu) {
2864#endif
2865		s = splnet();
2866		pfsync_sendout();
2867		splx(s);
2868
2869		nlen = sizeof(struct pfsync_subheader) + pfsync_qs[q].len;
2870	}
2871
2872	sc->sc_len += nlen;
2873	TAILQ_INSERT_TAIL(&sc->sc_qs[q], st, sync_list);
2874	st->sync_state = q;
2875}
2876
2877void
2878pfsync_q_del(struct pf_state *st)
2879{
2880#ifdef __FreeBSD__
2881	struct pfsync_softc *sc = V_pfsyncif;
2882#else
2883	struct pfsync_softc *sc = pfsyncif;
2884#endif
2885	int q = st->sync_state;
2886
2887#ifdef __FreeBSD__
2888	KASSERT(st->sync_state != PFSYNC_S_NONE,
2889		("%s: st->sync_state != PFSYNC_S_NONE", __FUNCTION__));
2890#else
2891	KASSERT(st->sync_state != PFSYNC_S_NONE);
2892#endif
2893
2894	sc->sc_len -= pfsync_qs[q].len;
2895	TAILQ_REMOVE(&sc->sc_qs[q], st, sync_list);
2896	st->sync_state = PFSYNC_S_NONE;
2897
2898	if (TAILQ_EMPTY(&sc->sc_qs[q]))
2899		sc->sc_len -= sizeof(struct pfsync_subheader);
2900}
2901
2902#ifdef notyet
2903void
2904pfsync_update_tdb(struct tdb *t, int output)
2905{
2906#ifdef __FreeBSD__
2907	struct pfsync_softc *sc = V_pfsyncif;
2908#else
2909	struct pfsync_softc *sc = pfsyncif;
2910#endif
2911	size_t nlen = sizeof(struct pfsync_tdb);
2912	int s;
2913
2914	if (sc == NULL)
2915		return;
2916
2917	if (!ISSET(t->tdb_flags, TDBF_PFSYNC)) {
2918		if (TAILQ_EMPTY(&sc->sc_tdb_q))
2919			nlen += sizeof(struct pfsync_subheader);
2920
2921		if (sc->sc_len + nlen > sc->sc_if.if_mtu) {
2922			s = splnet();
2923			PF_LOCK();
2924			pfsync_sendout();
2925			PF_UNLOCK();
2926			splx(s);
2927
2928			nlen = sizeof(struct pfsync_subheader) +
2929			    sizeof(struct pfsync_tdb);
2930		}
2931
2932		sc->sc_len += nlen;
2933		TAILQ_INSERT_TAIL(&sc->sc_tdb_q, t, tdb_sync_entry);
2934		SET(t->tdb_flags, TDBF_PFSYNC);
2935		t->tdb_updates = 0;
2936	} else {
2937		if (++t->tdb_updates >= sc->sc_maxupdates)
2938			schednetisr(NETISR_PFSYNC);
2939	}
2940
2941	if (output)
2942		SET(t->tdb_flags, TDBF_PFSYNC_RPL);
2943	else
2944		CLR(t->tdb_flags, TDBF_PFSYNC_RPL);
2945}
2946
2947void
2948pfsync_delete_tdb(struct tdb *t)
2949{
2950#ifdef __FreeBSD__
2951	struct pfsync_softc *sc = V_pfsyncif;
2952#else
2953	struct pfsync_softc *sc = pfsyncif;
2954#endif
2955
2956	if (sc == NULL || !ISSET(t->tdb_flags, TDBF_PFSYNC))
2957		return;
2958
2959	sc->sc_len -= sizeof(struct pfsync_tdb);
2960	TAILQ_REMOVE(&sc->sc_tdb_q, t, tdb_sync_entry);
2961	CLR(t->tdb_flags, TDBF_PFSYNC);
2962
2963	if (TAILQ_EMPTY(&sc->sc_tdb_q))
2964		sc->sc_len -= sizeof(struct pfsync_subheader);
2965}
2966
2967int
2968pfsync_out_tdb(struct tdb *t, struct mbuf *m, int offset)
2969{
2970	struct pfsync_tdb *ut = (struct pfsync_tdb *)(m->m_data + offset);
2971
2972	bzero(ut, sizeof(*ut));
2973	ut->spi = t->tdb_spi;
2974	bcopy(&t->tdb_dst, &ut->dst, sizeof(ut->dst));
2975	/*
2976	 * When a failover happens, the master's rpl is probably above
2977	 * what we see here (we may be up to a second late), so
2978	 * increase it a bit for outbound tdbs to manage most such
2979	 * situations.
2980	 *
2981	 * For now, just add an offset that is likely to be larger
2982	 * than the number of packets we can see in one second. The RFC
2983	 * just says the next packet must have a higher seq value.
2984	 *
2985	 * XXX What is a good algorithm for this? We could use
2986	 * a rate-determined increase, but to know it, we would have
2987	 * to extend struct tdb.
2988	 * XXX pt->rpl can wrap over MAXINT, but if so the real tdb
2989	 * will soon be replaced anyway. For now, just don't handle
2990	 * this edge case.
2991	 */
2992#define RPL_INCR 16384
2993	ut->rpl = htonl(t->tdb_rpl + (ISSET(t->tdb_flags, TDBF_PFSYNC_RPL) ?
2994	    RPL_INCR : 0));
2995	ut->cur_bytes = htobe64(t->tdb_cur_bytes);
2996	ut->sproto = t->tdb_sproto;
2997
2998	return (sizeof(*ut));
2999}
3000#endif
3001
3002void
3003pfsync_bulk_start(void)
3004{
3005#ifdef __FreeBSD__
3006	struct pfsync_softc *sc = V_pfsyncif;
3007#else
3008	struct pfsync_softc *sc = pfsyncif;
3009#endif
3010
3011#ifdef __FreeBSD__
3012	if (V_pf_status.debug >= PF_DEBUG_MISC)
3013#else
3014	if (pf_status.debug >= PF_DEBUG_MISC)
3015#endif
3016		printf("pfsync: received bulk update request\n");
3017
3018#ifdef __FreeBSD__
3019	PF_LOCK();
3020	if (TAILQ_EMPTY(&V_state_list))
3021#else
3022	if (TAILQ_EMPTY(&state_list))
3023#endif
3024		pfsync_bulk_status(PFSYNC_BUS_END);
3025	else {
3026		sc->sc_ureq_received = time_uptime;
3027		if (sc->sc_bulk_next == NULL)
3028#ifdef __FreeBSD__
3029			sc->sc_bulk_next = TAILQ_FIRST(&V_state_list);
3030#else
3031			sc->sc_bulk_next = TAILQ_FIRST(&state_list);
3032#endif
3033			sc->sc_bulk_last = sc->sc_bulk_next;
3034
3035			pfsync_bulk_status(PFSYNC_BUS_START);
3036			callout_reset(&sc->sc_bulk_tmo, 1,
3037			    pfsync_bulk_update, sc);
3038	}
3039#ifdef __FreeBSD__
3040	PF_UNLOCK();
3041#endif
3042}
3043
3044void
3045pfsync_bulk_update(void *arg)
3046{
3047	struct pfsync_softc *sc = arg;
3048	struct pf_state *st = sc->sc_bulk_next;
3049	int i = 0;
3050	int s;
3051
3052	PF_LOCK_ASSERT();
3053
3054	s = splsoftnet();
3055#ifdef __FreeBSD__
3056	CURVNET_SET(sc->sc_ifp->if_vnet);
3057#endif
3058	for (;;) {
3059		if (st->sync_state == PFSYNC_S_NONE &&
3060		    st->timeout < PFTM_MAX &&
3061		    st->pfsync_time <= sc->sc_ureq_received) {
3062			pfsync_update_state_req(st);
3063			i++;
3064		}
3065
3066		st = TAILQ_NEXT(st, entry_list);
3067		if (st == NULL)
3068#ifdef __FreeBSD__
3069			st = TAILQ_FIRST(&V_state_list);
3070#else
3071			st = TAILQ_FIRST(&state_list);
3072#endif
3073
3074		if (st == sc->sc_bulk_last) {
3075			/* we're done */
3076			sc->sc_bulk_next = NULL;
3077			sc->sc_bulk_last = NULL;
3078			pfsync_bulk_status(PFSYNC_BUS_END);
3079			break;
3080		}
3081
3082#ifdef __FreeBSD__
3083		if (i > 1 && (sc->sc_ifp->if_mtu - sc->sc_len) <
3084#else
3085		if (i > 1 && (sc->sc_if.if_mtu - sc->sc_len) <
3086#endif
3087		    sizeof(struct pfsync_state)) {
3088			/* we've filled a packet */
3089			sc->sc_bulk_next = st;
3090#ifdef __FreeBSD__
3091			callout_reset(&sc->sc_bulk_tmo, 1,
3092			    pfsync_bulk_update, sc);
3093#else
3094			timeout_add(&sc->sc_bulk_tmo, 1);
3095#endif
3096			break;
3097		}
3098	}
3099
3100#ifdef __FreeBSD__
3101	CURVNET_RESTORE();
3102#endif
3103	splx(s);
3104}
3105
3106void
3107pfsync_bulk_status(u_int8_t status)
3108{
3109	struct {
3110		struct pfsync_subheader subh;
3111		struct pfsync_bus bus;
3112	} __packed r;
3113
3114#ifdef __FreeBSD__
3115	struct pfsync_softc *sc = V_pfsyncif;
3116#else
3117	struct pfsync_softc *sc = pfsyncif;
3118#endif
3119
3120	PF_LOCK_ASSERT();
3121
3122	bzero(&r, sizeof(r));
3123
3124	r.subh.action = PFSYNC_ACT_BUS;
3125	r.subh.count = htons(1);
3126
3127#ifdef __FreeBSD__
3128	r.bus.creatorid = V_pf_status.hostid;
3129#else
3130	r.bus.creatorid = pf_status.hostid;
3131#endif
3132	r.bus.endtime = htonl(time_uptime - sc->sc_ureq_received);
3133	r.bus.status = status;
3134
3135	pfsync_send_plus(&r, sizeof(r));
3136}
3137
3138void
3139pfsync_bulk_fail(void *arg)
3140{
3141	struct pfsync_softc *sc = arg;
3142
3143#ifdef __FreeBSD__
3144	CURVNET_SET(sc->sc_ifp->if_vnet);
3145#endif
3146
3147	if (sc->sc_bulk_tries++ < PFSYNC_MAX_BULKTRIES) {
3148		/* Try again */
3149#ifdef __FreeBSD__
3150		callout_reset(&sc->sc_bulkfail_tmo, 5 * hz,
3151		    pfsync_bulk_fail, V_pfsyncif);
3152#else
3153		timeout_add_sec(&sc->sc_bulkfail_tmo, 5);
3154#endif
3155		PF_LOCK();
3156		pfsync_request_update(0, 0);
3157		PF_UNLOCK();
3158	} else {
3159		/* Pretend like the transfer was ok */
3160		sc->sc_ureq_sent = 0;
3161		sc->sc_bulk_tries = 0;
3162#if NCARP > 0
3163#ifdef notyet
3164#ifdef __FreeBSD__
3165		if (!sc->pfsync_sync_ok)
3166#else
3167		if (!pfsync_sync_ok)
3168#endif
3169			carp_group_demote_adj(&sc->sc_if, -1);
3170#endif
3171#endif
3172#ifdef __FreeBSD__
3173		sc->pfsync_sync_ok = 1;
3174#else
3175		pfsync_sync_ok = 1;
3176#endif
3177#ifdef __FreeBSD__
3178		if (V_pf_status.debug >= PF_DEBUG_MISC)
3179#else
3180		if (pf_status.debug >= PF_DEBUG_MISC)
3181#endif
3182			printf("pfsync: failed to receive bulk update\n");
3183	}
3184
3185#ifdef __FreeBSD__
3186	CURVNET_RESTORE();
3187#endif
3188}
3189
3190void
3191pfsync_send_plus(void *plus, size_t pluslen)
3192{
3193#ifdef __FreeBSD__
3194	struct pfsync_softc *sc = V_pfsyncif;
3195#else
3196	struct pfsync_softc *sc = pfsyncif;
3197#endif
3198	int s;
3199
3200	PF_LOCK_ASSERT();
3201
3202#ifdef __FreeBSD__
3203	if (sc->sc_len + pluslen > sc->sc_ifp->if_mtu) {
3204#else
3205	if (sc->sc_len + pluslen > sc->sc_if.if_mtu) {
3206#endif
3207		s = splnet();
3208		pfsync_sendout();
3209		splx(s);
3210	}
3211
3212	sc->sc_plus = plus;
3213	sc->sc_len += (sc->sc_pluslen = pluslen);
3214
3215	s = splnet();
3216	pfsync_sendout();
3217	splx(s);
3218}
3219
3220int
3221pfsync_up(void)
3222{
3223#ifdef __FreeBSD__
3224	struct pfsync_softc *sc = V_pfsyncif;
3225#else
3226	struct pfsync_softc *sc = pfsyncif;
3227#endif
3228
3229#ifdef __FreeBSD__
3230	if (sc == NULL || !ISSET(sc->sc_ifp->if_flags, IFF_DRV_RUNNING))
3231#else
3232	if (sc == NULL || !ISSET(sc->sc_if.if_flags, IFF_RUNNING))
3233#endif
3234		return (0);
3235
3236	return (1);
3237}
3238
3239int
3240pfsync_state_in_use(struct pf_state *st)
3241{
3242#ifdef __FreeBSD__
3243	struct pfsync_softc *sc = V_pfsyncif;
3244#else
3245	struct pfsync_softc *sc = pfsyncif;
3246#endif
3247
3248	if (sc == NULL)
3249		return (0);
3250
3251	if (st->sync_state != PFSYNC_S_NONE ||
3252	    st == sc->sc_bulk_next ||
3253	    st == sc->sc_bulk_last)
3254		return (1);
3255
3256	return (0);
3257}
3258
3259u_int pfsync_ints;
3260u_int pfsync_tmos;
3261
3262void
3263pfsync_timeout(void *arg)
3264{
3265#if defined(__FreeBSD__) && defined(VIMAGE)
3266	struct pfsync_softc *sc = arg;
3267#endif
3268	int s;
3269
3270#ifdef __FreeBSD__
3271	CURVNET_SET(sc->sc_ifp->if_vnet);
3272#endif
3273
3274	pfsync_tmos++;
3275
3276	s = splnet();
3277#ifdef __FreeBSD__
3278	PF_LOCK();
3279#endif
3280	pfsync_sendout();
3281#ifdef __FreeBSD__
3282	PF_UNLOCK();
3283#endif
3284	splx(s);
3285
3286#ifdef __FreeBSD__
3287	CURVNET_RESTORE();
3288#endif
3289}
3290
3291/* this is a softnet/netisr handler */
3292void
3293#ifdef __FreeBSD__
3294pfsyncintr(void *arg)
3295{
3296	struct pfsync_softc *sc = arg;
3297	struct mbuf *m, *n;
3298
3299	CURVNET_SET(sc->sc_ifp->if_vnet);
3300	pfsync_ints++;
3301
3302	IF_DEQUEUE_ALL(&sc->sc_ifp->if_snd, m);
3303
3304	for (; m != NULL; m = n) {
3305
3306		n = m->m_nextpkt;
3307		m->m_nextpkt = NULL;
3308		if (ip_output(m, NULL, NULL, IP_RAWOUTPUT, &sc->sc_imo, NULL)
3309		    == 0)
3310			V_pfsyncstats.pfsyncs_opackets++;
3311		else
3312			V_pfsyncstats.pfsyncs_oerrors++;
3313	}
3314	CURVNET_RESTORE();
3315}
3316#else
3317pfsyncintr(void)
3318{
3319	int s;
3320
3321	pfsync_ints++;
3322
3323	s = splnet();
3324	pfsync_sendout();
3325	splx(s);
3326}
3327#endif
3328
3329int
3330pfsync_sysctl(int *name, u_int namelen, void *oldp, size_t *oldlenp, void *newp,
3331    size_t newlen)
3332{
3333
3334#ifdef notyet
3335	/* All sysctl names at this level are terminal. */
3336	if (namelen != 1)
3337		return (ENOTDIR);
3338
3339	switch (name[0]) {
3340	case PFSYNCCTL_STATS:
3341		if (newp != NULL)
3342			return (EPERM);
3343		return (sysctl_struct(oldp, oldlenp, newp, newlen,
3344		    &V_pfsyncstats, sizeof(V_pfsyncstats)));
3345	}
3346#endif
3347	return (ENOPROTOOPT);
3348}
3349
3350#ifdef __FreeBSD__
3351void
3352pfsync_ifdetach(void *arg, struct ifnet *ifp)
3353{
3354	struct pfsync_softc *sc = (struct pfsync_softc *)arg;
3355	struct ip_moptions *imo;
3356
3357	if (sc == NULL || sc->sc_sync_if != ifp)
3358		return;         /* not for us; unlocked read */
3359
3360	CURVNET_SET(sc->sc_ifp->if_vnet);
3361
3362	PF_LOCK();
3363
3364	/* Deal with a member interface going away from under us. */
3365	sc->sc_sync_if = NULL;
3366	imo = &sc->sc_imo;
3367	if (imo->imo_num_memberships > 0) {
3368		KASSERT(imo->imo_num_memberships == 1,
3369		    ("%s: imo_num_memberships != 1", __func__));
3370		/*
3371		 * Our event handler is always called after protocol
3372		 * domains have been detached from the underlying ifnet.
3373		 * Do not call in_delmulti(); we held a single reference
3374		 * which the protocol domain has purged in in_purgemaddrs().
3375		 */
3376		PF_UNLOCK();
3377		imo->imo_membership[--imo->imo_num_memberships] = NULL;
3378		PF_LOCK();
3379		imo->imo_multicast_ifp = NULL;
3380	}
3381
3382	PF_UNLOCK();
3383
3384	CURVNET_RESTORE();
3385}
3386
3387static int
3388vnet_pfsync_init(const void *unused)
3389{
3390	int error = 0;
3391
3392	pfsyncattach(0);
3393
3394	error = swi_add(NULL, "pfsync", pfsyncintr, V_pfsyncif,
3395		SWI_NET, INTR_MPSAFE, &pfsync_swi.pfsync_swi_cookie);
3396	if (error)
3397		panic("%s: swi_add %d", __func__, error);
3398
3399	PF_LOCK();
3400	pfsync_state_import_ptr = pfsync_state_import;
3401	pfsync_up_ptr = pfsync_up;
3402	pfsync_insert_state_ptr = pfsync_insert_state;
3403	pfsync_update_state_ptr = pfsync_update_state;
3404	pfsync_delete_state_ptr = pfsync_delete_state;
3405	pfsync_clear_states_ptr = pfsync_clear_states;
3406	pfsync_state_in_use_ptr = pfsync_state_in_use;
3407	pfsync_defer_ptr = pfsync_defer;
3408	PF_UNLOCK();
3409
3410	return (0);
3411}
3412
3413static int
3414vnet_pfsync_uninit(const void *unused)
3415{
3416
3417	swi_remove(pfsync_swi.pfsync_swi_cookie);
3418
3419	PF_LOCK();
3420	pfsync_state_import_ptr = NULL;
3421	pfsync_up_ptr = NULL;
3422	pfsync_insert_state_ptr = NULL;
3423	pfsync_update_state_ptr = NULL;
3424	pfsync_delete_state_ptr = NULL;
3425	pfsync_clear_states_ptr = NULL;
3426	pfsync_state_in_use_ptr = NULL;
3427	pfsync_defer_ptr = NULL;
3428	PF_UNLOCK();
3429
3430	if_clone_detach(&pfsync_cloner);
3431
3432	return (0);
3433}
3434
3435/* Define startup order. */
3436#define	PFSYNC_SYSINIT_ORDER	SI_SUB_PROTO_IF
3437#define	PFSYNC_MODEVENT_ORDER	(SI_ORDER_FIRST) /* On boot slot in here. */
3438#define	PFSYNC_VNET_ORDER	(PFSYNC_MODEVENT_ORDER + 2) /* Later still. */
3439
3440/*
3441 * Starting up.
3442 * VNET_SYSINIT is called for each existing vnet and each new vnet.
3443 */
3444VNET_SYSINIT(vnet_pfsync_init, PFSYNC_SYSINIT_ORDER, PFSYNC_VNET_ORDER,
3445    vnet_pfsync_init, NULL);
3446
3447/*
3448 * Closing up shop. These are done in REVERSE ORDER,
3449 * Not called on reboot.
3450 * VNET_SYSUNINIT is called for each exiting vnet as it exits.
3451 */
3452VNET_SYSUNINIT(vnet_pfsync_uninit, PFSYNC_SYSINIT_ORDER, PFSYNC_VNET_ORDER,
3453    vnet_pfsync_uninit, NULL);
3454static int
3455pfsync_modevent(module_t mod, int type, void *data)
3456{
3457	int error = 0;
3458
3459	switch (type) {
3460	case MOD_LOAD:
3461#ifndef __FreeBSD__
3462		pfsyncattach(0);
3463#endif
3464		break;
3465	case MOD_UNLOAD:
3466#ifndef __FreeBSD__
3467		if_clone_detach(&pfsync_cloner);
3468#endif
3469		break;
3470	default:
3471		error = EINVAL;
3472		break;
3473	}
3474
3475	return error;
3476}
3477
3478static moduledata_t pfsync_mod = {
3479	"pfsync",
3480	pfsync_modevent,
3481	0
3482};
3483
3484#define PFSYNC_MODVER 1
3485
3486DECLARE_MODULE(pfsync, pfsync_mod, SI_SUB_PSEUDO, SI_ORDER_ANY);
3487MODULE_VERSION(pfsync, PFSYNC_MODVER);
3488MODULE_DEPEND(pfsync, pf, PF_MODVER, PF_MODVER, PF_MODVER);
3489#endif /* __FreeBSD__ */
3490