1 /* $OpenBSD: if_pfsync.c,v 1.98 2008/06/29 08:42:15 mcbride Exp $ */
4 * Copyright (c) 2002 Michael Shalayeff
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
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.
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.
30 #include "opt_inet6.h"
33 #include <sys/param.h>
34 #include <sys/endian.h>
37 #include <sys/systm.h>
40 #include <sys/socket.h>
41 #include <sys/kernel.h>
42 #include <sys/malloc.h>
43 #include <sys/module.h>
44 #include <sys/sockio.h>
45 #include <sys/thread2.h>
47 #include <machine/inttypes.h>
50 #include <net/if_types.h>
51 #include <net/route.h>
53 #include <netinet/in.h>
54 #include <netinet/if_ether.h>
55 #include <netinet/ip_carp.h>
56 #include <netinet/tcp.h>
57 #include <netinet/tcp_seq.h>
60 #include <netinet/in_systm.h>
61 #include <netinet/in_var.h>
62 #include <netinet/ip.h>
63 #include <netinet/ip_var.h>
67 #include <netinet6/nd6.h>
70 #include <net/pf/pfvar.h>
71 #include <net/pf/if_pfsync.h>
73 #define PFSYNCNAME "pfsync"
75 #define PFSYNC_MINMTU \
76 (sizeof(struct pfsync_header) + sizeof(struct pf_state))
79 #define DPRINTF(x) do { if (pfsyncdebug) kprintf x ; } while (0)
85 struct pfsync_softc *pfsyncif = NULL;
86 struct pfsyncstats pfsyncstats;
88 void pfsyncattach(int);
89 static int pfsync_clone_destroy(struct ifnet *);
90 static int pfsync_clone_create(struct if_clone *, int, caddr_t);
91 void pfsync_setmtu(struct pfsync_softc *, int);
92 int pfsync_alloc_scrub_memory(struct pfsync_state_peer *,
93 struct pf_state_peer *);
94 int pfsyncoutput(struct ifnet *, struct mbuf *, struct sockaddr *,
96 int pfsyncioctl(struct ifnet *, u_long, caddr_t, struct ucred *);
97 void pfsyncstart(struct ifnet *);
99 struct mbuf *pfsync_get_mbuf(struct pfsync_softc *, u_int8_t, void **);
100 int pfsync_request_update(struct pfsync_state_upd *, struct in_addr *);
101 int pfsync_sendout(struct pfsync_softc *);
102 int pfsync_sendout_mbuf(struct pfsync_softc *, struct mbuf *);
103 void pfsync_timeout(void *);
104 void pfsync_send_bus(struct pfsync_softc *, u_int8_t);
105 void pfsync_bulk_update(void *);
106 void pfsync_bulkfail(void *);
108 static MALLOC_DEFINE(M_PFSYNC, PFSYNCNAME, "Packet Filter State Sync. Interface");
109 static LIST_HEAD(pfsync_list, pfsync_softc) pfsync_list;
113 struct if_clone pfsync_cloner =
114 IF_CLONE_INITIALIZER("pfsync", pfsync_clone_create, pfsync_clone_destroy, 1 ,1);
117 pfsyncattach(int npfsync)
119 if_clone_attach(&pfsync_cloner);
122 pfsync_clone_create(struct if_clone *ifc, int unit, caddr_t param __unused)
124 struct pfsync_softc *sc;
127 lwkt_gettoken(&pf_token);
129 MALLOC(sc, struct pfsync_softc *, sizeof(*sc), M_PFSYNC,
133 sc->sc_mbuf_net = NULL;
134 sc->sc_mbuf_tdb = NULL;
135 sc->sc_statep.s = NULL;
136 sc->sc_statep_net.s = NULL;
137 sc->sc_statep_tdb.t = NULL;
138 sc->sc_maxupdates = 128;
139 sc->sc_sync_peer.s_addr =htonl(INADDR_PFSYNC_GROUP);
140 sc->sc_sendaddr.s_addr = htonl(INADDR_PFSYNC_GROUP);
141 sc->sc_ureq_received = 0;
142 sc->sc_ureq_sent = 0;
143 sc->sc_bulk_send_next = NULL;
144 sc->sc_bulk_terminator = NULL;
145 sc->sc_imo.imo_max_memberships = IP_MAX_MEMBERSHIPS;
146 lwkt_reltoken(&pf_token);
148 ksnprintf(ifp->if_xname, sizeof ifp->if_xname, "pfsync%d", unit);
149 if_initname(ifp, ifc->ifc_name, unit);
150 ifp->if_ioctl = pfsyncioctl;
151 ifp->if_output = pfsyncoutput;
152 ifp->if_start = pfsyncstart;
153 ifp->if_type = IFT_PFSYNC;
154 ifp->if_snd.ifq_maxlen = ifqmaxlen;
155 ifp->if_hdrlen = PFSYNC_HDRLEN;
156 ifp->if_baudrate = IF_Mbps(100);
158 pfsync_setmtu(sc, MCLBYTES);
159 callout_init(&sc->sc_tmo);
160 /* callout_init(&sc->sc_tdb_tmo); XXX we don't support tdb (yet) */
161 callout_init(&sc->sc_bulk_tmo);
162 callout_init(&sc->sc_bulkfail_tmo);
163 if_attach(ifp, NULL);
165 LIST_INSERT_HEAD(&pfsync_list, sc, sc_next);
169 if_addgroup(ifp, "carp");
173 bpfattach(&sc->sc_if, DLT_PFSYNC, PFSYNC_HDRLEN);
175 lwkt_gettoken(&pf_token);
177 lwkt_reltoken(&pf_token);
182 pfsync_clone_destroy(struct ifnet *ifp)
184 lwkt_gettoken(&pf_token);
185 lwkt_reltoken(&pf_token);
187 struct pfsync_softc *sc = ifp->if_softc;
188 callout_stop(&sc->sc_tmo);
189 /* callout_stop(&sc->sc_tdb_tmo); XXX we don't support tdb (yet) */
190 callout_stop(&sc->sc_bulk_tmo);
191 callout_stop(&sc->sc_bulkfail_tmo);
194 carp_group_demote_adj(&sc->sc_if, -1);
200 lwkt_gettoken(&pf_token);
201 LIST_REMOVE(sc, sc_next);
203 lwkt_reltoken(&pf_token);
210 * Start output on the pfsync interface.
213 pfsyncstart(struct ifnet *ifp)
216 IF_DROP(&ifp->if_snd);
217 IF_DRAIN(&ifp->if_snd);
222 pfsync_alloc_scrub_memory(struct pfsync_state_peer *s,
223 struct pf_state_peer *d)
225 if (s->scrub.scrub_flag && d->scrub == NULL) {
226 d->scrub = kmalloc(sizeof(struct pf_state_scrub), M_PFSYNC, M_NOWAIT|M_ZERO);
228 if (d->scrub == NULL)
236 pfsync_state_export(struct pfsync_state *sp, struct pf_state *st)
238 bzero(sp, sizeof(struct pfsync_state));
240 /* copy from state key */
241 sp->key[PF_SK_WIRE].addr[0] = st->key[PF_SK_WIRE]->addr[0];
242 sp->key[PF_SK_WIRE].addr[1] = st->key[PF_SK_WIRE]->addr[1];
243 sp->key[PF_SK_WIRE].port[0] = st->key[PF_SK_WIRE]->port[0];
244 sp->key[PF_SK_WIRE].port[1] = st->key[PF_SK_WIRE]->port[1];
245 sp->key[PF_SK_STACK].addr[0] = st->key[PF_SK_STACK]->addr[0];
246 sp->key[PF_SK_STACK].addr[1] = st->key[PF_SK_STACK]->addr[1];
247 sp->key[PF_SK_STACK].port[0] = st->key[PF_SK_STACK]->port[0];
248 sp->key[PF_SK_STACK].port[1] = st->key[PF_SK_STACK]->port[1];
249 sp->proto = st->key[PF_SK_WIRE]->proto;
250 sp->af = st->key[PF_SK_WIRE]->af;
252 /* copy from state */
253 strlcpy(sp->ifname, st->kif->pfik_name, sizeof(sp->ifname));
254 bcopy(&st->rt_addr, &sp->rt_addr, sizeof(sp->rt_addr));
255 sp->creation = htonl(time_second - st->creation);
256 sp->expire = pf_state_expires(st);
257 if (sp->expire <= time_second)
258 sp->expire = htonl(0);
260 sp->expire = htonl(sp->expire - time_second);
262 sp->direction = st->direction;
264 sp->timeout = st->timeout;
265 sp->state_flags = st->state_flags;
267 sp->sync_flags |= PFSYNC_FLAG_SRCNODE;
268 if (st->nat_src_node)
269 sp->sync_flags |= PFSYNC_FLAG_NATSRCNODE;
271 bcopy(&st->id, &sp->id, sizeof(sp->id));
272 sp->creatorid = st->creatorid;
273 pf_state_peer_hton(&st->src, &sp->src);
274 pf_state_peer_hton(&st->dst, &sp->dst);
276 if (st->rule.ptr == NULL)
277 sp->rule = htonl(-1);
279 sp->rule = htonl(st->rule.ptr->nr);
280 if (st->anchor.ptr == NULL)
281 sp->anchor = htonl(-1);
283 sp->anchor = htonl(st->anchor.ptr->nr);
284 if (st->nat_rule.ptr == NULL)
285 sp->nat_rule = htonl(-1);
287 sp->nat_rule = htonl(st->nat_rule.ptr->nr);
289 pf_state_counter_hton(st->packets[0], sp->packets[0]);
290 pf_state_counter_hton(st->packets[1], sp->packets[1]);
291 pf_state_counter_hton(st->bytes[0], sp->bytes[0]);
292 pf_state_counter_hton(st->bytes[1], sp->bytes[1]);
297 pfsync_state_import(struct pfsync_state *sp, u_int8_t flags)
299 struct pf_state *st = NULL;
300 struct pf_state_key *skw = NULL, *sks = NULL;
301 struct pf_rule *r = NULL;
306 if (sp->creatorid == 0 && pf_status.debug >= PF_DEBUG_MISC) {
307 kprintf("pfsync_insert_net_state: invalid creator id:"
308 " %08x\n", ntohl(sp->creatorid));
312 if ((kif = pfi_kif_get(sp->ifname)) == NULL) {
313 if (pf_status.debug >= PF_DEBUG_MISC)
314 kprintf("pfsync_insert_net_state: "
315 "unknown interface: %s\n", sp->ifname);
316 if (flags & PFSYNC_SI_IOCTL)
318 return (0); /* skip this state */
322 * If the ruleset checksums match or the state is coming from the ioctl,
323 * it's safe to associate the state with the rule of that number.
325 if (sp->rule != htonl(-1) && sp->anchor == htonl(-1) &&
326 (flags & (PFSYNC_SI_IOCTL | PFSYNC_SI_CKSUM)) && ntohl(sp->rule) <
327 pf_main_ruleset.rules[PF_RULESET_FILTER].active.rcount)
328 r = pf_main_ruleset.rules[
329 PF_RULESET_FILTER].active.ptr_array[ntohl(sp->rule)];
331 r = &pf_default_rule;
333 if ((r->max_states && r->states_cur >= r->max_states))
336 if (flags & PFSYNC_SI_IOCTL)
337 pool_flags = M_WAITOK | M_NULLOK | M_ZERO;
339 pool_flags = M_WAITOK | M_ZERO;
341 if ((st = kmalloc(sizeof(struct pf_state), M_PFSYNC, pool_flags)) == NULL)
344 if ((skw = pf_alloc_state_key(pool_flags)) == NULL)
347 if (PF_ANEQ(&sp->key[PF_SK_WIRE].addr[0],
348 &sp->key[PF_SK_STACK].addr[0], sp->af) ||
349 PF_ANEQ(&sp->key[PF_SK_WIRE].addr[1],
350 &sp->key[PF_SK_STACK].addr[1], sp->af) ||
351 sp->key[PF_SK_WIRE].port[0] != sp->key[PF_SK_STACK].port[0] ||
352 sp->key[PF_SK_WIRE].port[1] != sp->key[PF_SK_STACK].port[1]) {
353 if ((sks = pf_alloc_state_key(pool_flags)) == NULL)
358 /* allocate memory for scrub info */
359 if (pfsync_alloc_scrub_memory(&sp->src, &st->src) ||
360 pfsync_alloc_scrub_memory(&sp->dst, &st->dst))
363 /* copy to state key(s) */
364 skw->addr[0] = sp->key[PF_SK_WIRE].addr[0];
365 skw->addr[1] = sp->key[PF_SK_WIRE].addr[1];
366 skw->port[0] = sp->key[PF_SK_WIRE].port[0];
367 skw->port[1] = sp->key[PF_SK_WIRE].port[1];
368 skw->proto = sp->proto;
371 sks->addr[0] = sp->key[PF_SK_STACK].addr[0];
372 sks->addr[1] = sp->key[PF_SK_STACK].addr[1];
373 sks->port[0] = sp->key[PF_SK_STACK].port[0];
374 sks->port[1] = sp->key[PF_SK_STACK].port[1];
375 sks->proto = sp->proto;
380 bcopy(&sp->rt_addr, &st->rt_addr, sizeof(st->rt_addr));
381 st->creation = time_second - ntohl(sp->creation);
382 st->expire = time_second;
384 /* XXX No adaptive scaling. */
385 st->expire -= r->timeout[sp->timeout] - ntohl(sp->expire);
388 st->expire = ntohl(sp->expire) + time_second;
389 st->direction = sp->direction;
391 st->timeout = sp->timeout;
392 st->state_flags = sp->state_flags;
393 if (!(flags & PFSYNC_SI_IOCTL))
394 st->sync_flags = PFSTATE_FROMSYNC;
396 bcopy(sp->id, &st->id, sizeof(st->id));
397 st->creatorid = sp->creatorid;
398 pf_state_peer_ntoh(&sp->src, &st->src);
399 pf_state_peer_ntoh(&sp->dst, &st->dst);
402 st->nat_rule.ptr = NULL;
403 st->anchor.ptr = NULL;
409 /* XXX when we have nat_rule/anchors, use STATE_INC_COUNTERS */
413 if ((error = pf_state_insert(kif, skw, sks, st)) != 0) {
414 /* XXX when we have nat_rule/anchors, use STATE_DEC_COUNTERS */
426 kfree(skw, M_PFSYNC);
428 kfree(sks, M_PFSYNC);
430 cleanup_state: /* pf_state_insert frees the state keys */
433 kfree(st->dst.scrub, M_PFSYNC);
435 kfree(st->src.scrub, M_PFSYNC);
442 pfsync_input(struct mbuf *m, ...)
444 struct ip *ip = mtod(m, struct ip *);
445 struct pfsync_header *ph;
446 struct pfsync_softc *sc = pfsyncif;
448 struct pf_state_key *sk;
449 struct pf_state_item *si;
450 struct pf_state_cmp id_key;
451 struct pfsync_state *sp;
452 struct pfsync_state_upd *up;
453 struct pfsync_state_del *dp;
454 struct pfsync_state_clr *cp;
455 struct pfsync_state_upd_req *rup;
456 struct pfsync_state_bus *bus;
458 struct pfsync_tdb *pt;
462 int iplen, action, error, i, count, offp, sfail, stale = 0;
465 /* This function is not yet called from anywhere */
466 /* Still we assume for safety that pf_token must be held */
467 ASSERT_LWKT_TOKEN_HELD(&pf_token);
469 pfsyncstats.pfsyncs_ipackets++;
471 /* verify that we have a sync interface configured */
472 if (!sc || !sc->sc_sync_ifp || !pf_status.running)
475 /* verify that the packet came in on the right interface */
476 if (sc->sc_sync_ifp != m->m_pkthdr.rcvif) {
477 pfsyncstats.pfsyncs_badif++;
481 /* verify that the IP TTL is 255. */
482 if (ip->ip_ttl != PFSYNC_DFLTTL) {
483 pfsyncstats.pfsyncs_badttl++;
487 iplen = ip->ip_hl << 2;
489 if (m->m_pkthdr.len < iplen + sizeof(*ph)) {
490 pfsyncstats.pfsyncs_hdrops++;
494 if (iplen + sizeof(*ph) > m->m_len) {
495 if ((m = m_pullup(m, iplen + sizeof(*ph))) == NULL) {
496 pfsyncstats.pfsyncs_hdrops++;
499 ip = mtod(m, struct ip *);
501 ph = (struct pfsync_header *)((char *)ip + iplen);
503 /* verify the version */
504 if (ph->version != PFSYNC_VERSION) {
505 pfsyncstats.pfsyncs_badver++;
512 /* make sure it's a valid action code */
513 if (action >= PFSYNC_ACT_MAX) {
514 pfsyncstats.pfsyncs_badact++;
518 /* Cheaper to grab this now than having to mess with mbufs later */
521 if (!bcmp(&ph->pf_chksum, &pf_status.pf_chksum, PF_MD5_DIGEST_LENGTH))
522 flags |= PFSYNC_SI_CKSUM;
525 case PFSYNC_ACT_CLR: {
526 struct pf_state *nexts;
527 struct pf_state_key *nextsk;
530 if ((mp = m_pulldown(m, iplen + sizeof(*ph),
531 sizeof(*cp), &offp)) == NULL) {
532 pfsyncstats.pfsyncs_badlen++;
535 cp = (struct pfsync_state_clr *)(mp->m_data + offp);
536 creatorid = cp->creatorid;
539 if (cp->ifname[0] == '\0') {
540 for (st = RB_MIN(pf_state_tree_id, &tree_id);
542 nexts = RB_NEXT(pf_state_tree_id, &tree_id, st);
543 if (st->creatorid == creatorid) {
544 st->sync_flags |= PFSTATE_FROMSYNC;
549 if ((kif = pfi_kif_get(cp->ifname)) == NULL) {
554 for (sk = RB_MIN(pf_state_tree,
555 &pf_statetbl); sk; sk = nextsk) {
556 nextsk = RB_NEXT(pf_state_tree,
558 TAILQ_FOREACH(si, &sk->states, entry) {
559 if (si->s->creatorid == creatorid) {
562 pf_unlink_state(si->s);
572 if ((mp = m_pulldown(m, iplen + sizeof(*ph),
573 count * sizeof(*sp), &offp)) == NULL) {
574 pfsyncstats.pfsyncs_badlen++;
579 for (i = 0, sp = (struct pfsync_state *)(mp->m_data + offp);
580 i < count; i++, sp++) {
581 /* check for invalid values */
582 if (sp->timeout >= PFTM_MAX ||
583 sp->src.state > PF_TCPS_PROXY_DST ||
584 sp->dst.state > PF_TCPS_PROXY_DST ||
585 sp->direction > PF_OUT ||
586 (sp->af != AF_INET && sp->af != AF_INET6)) {
587 if (pf_status.debug >= PF_DEBUG_MISC)
588 kprintf("pfsync_insert: PFSYNC_ACT_INS: "
590 pfsyncstats.pfsyncs_badval++;
594 if ((error = pfsync_state_import(sp, flags))) {
595 if (error == ENOMEM) {
605 if ((mp = m_pulldown(m, iplen + sizeof(*ph),
606 count * sizeof(*sp), &offp)) == NULL) {
607 pfsyncstats.pfsyncs_badlen++;
612 for (i = 0, sp = (struct pfsync_state *)(mp->m_data + offp);
613 i < count; i++, sp++) {
614 int flags = PFSYNC_FLAG_STALE;
616 /* check for invalid values */
617 if (sp->timeout >= PFTM_MAX ||
618 sp->src.state > PF_TCPS_PROXY_DST ||
619 sp->dst.state > PF_TCPS_PROXY_DST) {
620 if (pf_status.debug >= PF_DEBUG_MISC)
621 kprintf("pfsync_insert: PFSYNC_ACT_UPD: "
623 pfsyncstats.pfsyncs_badval++;
627 bcopy(sp->id, &id_key.id, sizeof(id_key.id));
628 id_key.creatorid = sp->creatorid;
630 st = pf_find_state_byid(&id_key);
632 /* insert the update */
633 if (pfsync_state_import(sp, flags))
634 pfsyncstats.pfsyncs_badstate++;
637 sk = st->key[PF_SK_WIRE]; /* XXX right one? */
639 if (sk->proto == IPPROTO_TCP) {
641 * The state should never go backwards except
642 * for syn-proxy states. Neither should the
643 * sequence window slide backwards.
645 if (st->src.state > sp->src.state &&
646 (st->src.state < PF_TCPS_PROXY_SRC ||
647 sp->src.state >= PF_TCPS_PROXY_SRC))
649 else if (SEQ_GT(st->src.seqlo,
650 ntohl(sp->src.seqlo)))
652 else if (st->dst.state > sp->dst.state) {
653 /* There might still be useful
654 * information about the src state here,
655 * so import that part of the update,
656 * then "fail" so we send the updated
657 * state back to the peer who is missing
658 * our what we know. */
659 pf_state_peer_ntoh(&sp->src, &st->src);
660 /* XXX do anything with timeouts? */
663 } else if (st->dst.state >= TCPS_SYN_SENT &&
664 SEQ_GT(st->dst.seqlo, ntohl(sp->dst.seqlo)))
668 * Non-TCP protocol state machine always go
671 if (st->src.state > sp->src.state)
673 else if (st->dst.state > sp->dst.state)
677 if (pf_status.debug >= PF_DEBUG_MISC)
678 kprintf("pfsync: %s stale update "
681 (sfail < 7 ? "ignoring"
683 (uintmax_t)be64toh(st->id),
684 ntohl(st->creatorid));
685 pfsyncstats.pfsyncs_stale++;
687 if (!(sp->sync_flags & PFSTATE_STALE)) {
688 /* we have a better state, send it */
689 if (sc->sc_mbuf != NULL && !stale)
694 PFSYNC_ACT_UPD, st, flags);
698 pfsync_alloc_scrub_memory(&sp->dst, &st->dst);
699 pf_state_peer_ntoh(&sp->src, &st->src);
700 pf_state_peer_ntoh(&sp->dst, &st->dst);
701 st->expire = ntohl(sp->expire) + time_second;
702 st->timeout = sp->timeout;
704 if (stale && sc->sc_mbuf != NULL)
709 * It's not strictly necessary for us to support the "uncompressed"
710 * delete action, but it's relatively simple and maintains consistency.
713 if ((mp = m_pulldown(m, iplen + sizeof(*ph),
714 count * sizeof(*sp), &offp)) == NULL) {
715 pfsyncstats.pfsyncs_badlen++;
720 for (i = 0, sp = (struct pfsync_state *)(mp->m_data + offp);
721 i < count; i++, sp++) {
722 bcopy(sp->id, &id_key.id, sizeof(id_key.id));
723 id_key.creatorid = sp->creatorid;
725 st = pf_find_state_byid(&id_key);
727 pfsyncstats.pfsyncs_badstate++;
730 st->sync_flags |= PFSTATE_FROMSYNC;
735 case PFSYNC_ACT_UPD_C: {
736 int update_requested = 0;
738 if ((mp = m_pulldown(m, iplen + sizeof(*ph),
739 count * sizeof(*up), &offp)) == NULL) {
740 pfsyncstats.pfsyncs_badlen++;
745 for (i = 0, up = (struct pfsync_state_upd *)(mp->m_data + offp);
746 i < count; i++, up++) {
747 /* check for invalid values */
748 if (up->timeout >= PFTM_MAX ||
749 up->src.state > PF_TCPS_PROXY_DST ||
750 up->dst.state > PF_TCPS_PROXY_DST) {
751 if (pf_status.debug >= PF_DEBUG_MISC)
752 kprintf("pfsync_insert: "
755 pfsyncstats.pfsyncs_badval++;
759 bcopy(up->id, &id_key.id, sizeof(id_key.id));
760 id_key.creatorid = up->creatorid;
762 st = pf_find_state_byid(&id_key);
764 /* We don't have this state. Ask for it. */
765 error = pfsync_request_update(up, &src);
766 if (error == ENOMEM) {
770 update_requested = 1;
771 pfsyncstats.pfsyncs_badstate++;
774 sk = st->key[PF_SK_WIRE]; /* XXX right one? */
776 if (sk->proto == IPPROTO_TCP) {
778 * The state should never go backwards except
779 * for syn-proxy states. Neither should the
780 * sequence window slide backwards.
782 if (st->src.state > up->src.state &&
783 (st->src.state < PF_TCPS_PROXY_SRC ||
784 up->src.state >= PF_TCPS_PROXY_SRC))
786 else if (st->dst.state > up->dst.state)
788 else if (SEQ_GT(st->src.seqlo,
789 ntohl(up->src.seqlo)))
791 else if (st->dst.state >= TCPS_SYN_SENT &&
792 SEQ_GT(st->dst.seqlo, ntohl(up->dst.seqlo)))
796 * Non-TCP protocol state machine always go
799 if (st->src.state > up->src.state)
801 else if (st->dst.state > up->dst.state)
805 if (pf_status.debug >= PF_DEBUG_MISC)
806 kprintf("pfsync: ignoring stale update "
807 "(%d) id: %016" PRIx64 " "
808 "creatorid: %08x\n", sfail,
810 ntohl(st->creatorid));
811 pfsyncstats.pfsyncs_stale++;
813 /* we have a better state, send it out */
814 if ((!stale || update_requested) &&
815 sc->sc_mbuf != NULL) {
817 update_requested = 0;
821 pfsync_pack_state(PFSYNC_ACT_UPD, st,
825 pfsync_alloc_scrub_memory(&up->dst, &st->dst);
826 pf_state_peer_ntoh(&up->src, &st->src);
827 pf_state_peer_ntoh(&up->dst, &st->dst);
828 st->expire = ntohl(up->expire) + time_second;
829 st->timeout = up->timeout;
831 if ((update_requested || stale) && sc->sc_mbuf)
836 case PFSYNC_ACT_DEL_C:
837 if ((mp = m_pulldown(m, iplen + sizeof(*ph),
838 count * sizeof(*dp), &offp)) == NULL) {
839 pfsyncstats.pfsyncs_badlen++;
844 for (i = 0, dp = (struct pfsync_state_del *)(mp->m_data + offp);
845 i < count; i++, dp++) {
846 bcopy(dp->id, &id_key.id, sizeof(id_key.id));
847 id_key.creatorid = dp->creatorid;
849 st = pf_find_state_byid(&id_key);
851 pfsyncstats.pfsyncs_badstate++;
854 st->sync_flags |= PFSTATE_FROMSYNC;
859 case PFSYNC_ACT_INS_F:
860 case PFSYNC_ACT_DEL_F:
861 /* not implemented */
863 case PFSYNC_ACT_UREQ:
864 if ((mp = m_pulldown(m, iplen + sizeof(*ph),
865 count * sizeof(*rup), &offp)) == NULL) {
866 pfsyncstats.pfsyncs_badlen++;
871 if (sc->sc_mbuf != NULL)
874 rup = (struct pfsync_state_upd_req *)(mp->m_data + offp);
875 i < count; i++, rup++) {
876 bcopy(rup->id, &id_key.id, sizeof(id_key.id));
877 id_key.creatorid = rup->creatorid;
879 if (id_key.id == 0 && id_key.creatorid == 0) {
880 sc->sc_ureq_received = mycpu->gd_time_seconds;
881 if (sc->sc_bulk_send_next == NULL)
882 sc->sc_bulk_send_next =
883 TAILQ_FIRST(&state_list);
884 sc->sc_bulk_terminator = sc->sc_bulk_send_next;
885 if (pf_status.debug >= PF_DEBUG_MISC)
886 kprintf("pfsync: received "
887 "bulk update request\n");
888 pfsync_send_bus(sc, PFSYNC_BUS_START);
889 lwkt_reltoken(&pf_token);
890 callout_init(&sc->sc_bulk_tmo);
891 lwkt_gettoken(&pf_token);
893 st = pf_find_state_byid(&id_key);
895 pfsyncstats.pfsyncs_badstate++;
899 pfsync_pack_state(PFSYNC_ACT_UPD,
903 if (sc->sc_mbuf != NULL)
908 /* If we're not waiting for a bulk update, who cares. */
909 if (sc->sc_ureq_sent == 0)
912 if ((mp = m_pulldown(m, iplen + sizeof(*ph),
913 sizeof(*bus), &offp)) == NULL) {
914 pfsyncstats.pfsyncs_badlen++;
917 bus = (struct pfsync_state_bus *)(mp->m_data + offp);
918 switch (bus->status) {
919 case PFSYNC_BUS_START:
920 lwkt_reltoken(&pf_token);
921 callout_reset(&sc->sc_bulkfail_tmo,
922 pf_pool_limits[PF_LIMIT_STATES].limit /
923 (PFSYNC_BULKPACKETS * sc->sc_maxcount),
924 pfsync_bulkfail, LIST_FIRST(&pfsync_list));
925 lwkt_gettoken(&pf_token);
926 if (pf_status.debug >= PF_DEBUG_MISC)
927 kprintf("pfsync: received bulk "
931 if (mycpu->gd_time_seconds - ntohl(bus->endtime) >=
933 /* that's it, we're happy */
934 sc->sc_ureq_sent = 0;
935 sc->sc_bulk_tries = 0;
936 lwkt_reltoken(&pf_token);
937 callout_stop(&sc->sc_bulkfail_tmo);
938 lwkt_gettoken(&pf_token);
940 if (!pfsync_sync_ok) {
941 lwkt_reltoken(&pf_token);
942 carp_group_demote_adj(&sc->sc_if, -1);
943 lwkt_gettoken(&pf_token);
947 if (pf_status.debug >= PF_DEBUG_MISC)
948 kprintf("pfsync: received valid "
949 "bulk update end\n");
951 if (pf_status.debug >= PF_DEBUG_MISC)
952 kprintf("pfsync: received invalid "
953 "bulk update end: bad timestamp\n");
959 case PFSYNC_ACT_TDB_UPD:
960 if ((mp = m_pulldown(m, iplen + sizeof(*ph),
961 count * sizeof(*pt), &offp)) == NULL) {
962 pfsyncstats.pfsyncs_badlen++;
966 for (i = 0, pt = (struct pfsync_tdb *)(mp->m_data + offp);
967 i < count; i++, pt++)
968 pfsync_update_net_tdb(pt);
980 pfsyncoutput(struct ifnet *ifp, struct mbuf *m, struct sockaddr *dst,
989 pfsyncioctl(struct ifnet *ifp, u_long cmd, caddr_t data, struct ucred *cr)
991 struct pfsync_softc *sc = ifp->if_softc;
992 struct ifreq *ifr = (struct ifreq *)data;
993 struct ip_moptions *imo = &sc->sc_imo;
994 struct pfsyncreq pfsyncr;
998 lwkt_gettoken(&pf_token);
1003 case SIOCSIFDSTADDR:
1005 if (ifp->if_flags & IFF_UP)
1006 ifp->if_flags |= IFF_RUNNING;
1008 ifp->if_flags &= ~IFF_RUNNING;
1011 if (ifr->ifr_mtu < PFSYNC_MINMTU) {
1012 lwkt_reltoken(&pf_token);
1015 if (ifr->ifr_mtu > MCLBYTES)
1016 ifr->ifr_mtu = MCLBYTES;
1018 if (ifr->ifr_mtu < ifp->if_mtu)
1020 pfsync_setmtu(sc, ifr->ifr_mtu);
1024 bzero(&pfsyncr, sizeof(pfsyncr));
1025 if (sc->sc_sync_ifp)
1026 strlcpy(pfsyncr.pfsyncr_syncdev,
1027 sc->sc_sync_ifp->if_xname, IFNAMSIZ);
1028 pfsyncr.pfsyncr_syncpeer = sc->sc_sync_peer;
1029 pfsyncr.pfsyncr_maxupdates = sc->sc_maxupdates;
1030 lwkt_reltoken(&pf_token);
1031 if ((error = copyout(&pfsyncr, ifr->ifr_data, sizeof(pfsyncr))))
1033 lwkt_gettoken(&pf_token);
1036 if ((error = priv_check_cred(cr, PRIV_ROOT, NULL_CRED_OKAY)) != 0) {
1037 lwkt_reltoken(&pf_token);
1040 if ((error = copyin(ifr->ifr_data, &pfsyncr, sizeof(pfsyncr)))) {
1041 lwkt_reltoken(&pf_token);
1045 if (pfsyncr.pfsyncr_syncpeer.s_addr == 0)
1046 sc->sc_sync_peer.s_addr = INADDR_PFSYNC_GROUP;
1048 sc->sc_sync_peer.s_addr =
1049 pfsyncr.pfsyncr_syncpeer.s_addr;
1051 if (pfsyncr.pfsyncr_maxupdates > 255) {
1052 lwkt_reltoken(&pf_token);
1055 sc->sc_maxupdates = pfsyncr.pfsyncr_maxupdates;
1057 if (pfsyncr.pfsyncr_syncdev[0] == 0) {
1058 sc->sc_sync_ifp = NULL;
1059 if (sc->sc_mbuf_net != NULL) {
1060 /* Don't keep stale pfsync packets around. */
1062 m_freem(sc->sc_mbuf_net);
1063 sc->sc_mbuf_net = NULL;
1064 sc->sc_statep_net.s = NULL;
1067 if (imo->imo_num_memberships > 0) {
1068 in_delmulti(imo->imo_membership[--imo->imo_num_memberships]);
1069 imo->imo_multicast_ifp = NULL;
1074 if ((sifp = ifunit(pfsyncr.pfsyncr_syncdev)) == NULL) {
1075 lwkt_reltoken(&pf_token);
1080 if (sifp->if_mtu < sc->sc_if.if_mtu ||
1081 (sc->sc_sync_ifp != NULL &&
1082 sifp->if_mtu < sc->sc_sync_ifp->if_mtu) ||
1083 sifp->if_mtu < MCLBYTES - sizeof(struct ip))
1085 sc->sc_sync_ifp = sifp;
1087 pfsync_setmtu(sc, sc->sc_if.if_mtu);
1089 if (imo->imo_num_memberships > 0) {
1090 in_delmulti(imo->imo_membership[--imo->imo_num_memberships]);
1091 imo->imo_multicast_ifp = NULL;
1094 if (sc->sc_sync_ifp &&
1095 sc->sc_sync_peer.s_addr == INADDR_PFSYNC_GROUP) {
1096 struct in_addr addr;
1098 if (!(sc->sc_sync_ifp->if_flags & IFF_MULTICAST)) {
1099 sc->sc_sync_ifp = NULL;
1100 lwkt_reltoken(&pf_token);
1102 return (EADDRNOTAVAIL);
1105 addr.s_addr = INADDR_PFSYNC_GROUP;
1107 if ((imo->imo_membership[0] =
1108 in_addmulti(&addr, sc->sc_sync_ifp)) == NULL) {
1109 sc->sc_sync_ifp = NULL;
1110 lwkt_reltoken(&pf_token);
1114 imo->imo_num_memberships++;
1115 imo->imo_multicast_ifp = sc->sc_sync_ifp;
1116 imo->imo_multicast_ttl = PFSYNC_DFLTTL;
1117 imo->imo_multicast_loop = 0;
1120 if (sc->sc_sync_ifp ||
1121 sc->sc_sendaddr.s_addr != INADDR_PFSYNC_GROUP) {
1122 /* Request a full state table update. */
1123 sc->sc_ureq_sent = mycpu->gd_time_seconds;
1126 carp_group_demote_adj(&sc->sc_if, 1);
1129 if (pf_status.debug >= PF_DEBUG_MISC)
1130 kprintf("pfsync: requesting bulk update\n");
1131 lwkt_reltoken(&pf_token);
1132 callout_reset(&sc->sc_bulkfail_tmo, 5 * hz,
1133 pfsync_bulkfail, LIST_FIRST(&pfsync_list));
1134 lwkt_gettoken(&pf_token);
1135 error = pfsync_request_update(NULL, NULL);
1136 if (error == ENOMEM) {
1137 lwkt_reltoken(&pf_token);
1148 lwkt_reltoken(&pf_token);
1152 lwkt_reltoken(&pf_token);
1157 pfsync_setmtu(struct pfsync_softc *sc, int mtu_req)
1161 if (sc->sc_sync_ifp && sc->sc_sync_ifp->if_mtu < mtu_req)
1162 mtu = sc->sc_sync_ifp->if_mtu;
1166 sc->sc_maxcount = (mtu - sizeof(struct pfsync_header)) /
1167 sizeof(struct pfsync_state);
1168 if (sc->sc_maxcount > 254)
1169 sc->sc_maxcount = 254;
1170 sc->sc_if.if_mtu = sizeof(struct pfsync_header) +
1171 sc->sc_maxcount * sizeof(struct pfsync_state);
1175 pfsync_get_mbuf(struct pfsync_softc *sc, u_int8_t action, void **sp)
1177 struct pfsync_header *h;
1181 ASSERT_LWKT_TOKEN_HELD(&pf_token);
1183 MGETHDR(m, M_WAITOK, MT_DATA);
1185 sc->sc_if.if_oerrors++;
1190 case PFSYNC_ACT_CLR:
1191 len = sizeof(struct pfsync_header) +
1192 sizeof(struct pfsync_state_clr);
1194 case PFSYNC_ACT_UPD_C:
1195 len = (sc->sc_maxcount * sizeof(struct pfsync_state_upd)) +
1196 sizeof(struct pfsync_header);
1198 case PFSYNC_ACT_DEL_C:
1199 len = (sc->sc_maxcount * sizeof(struct pfsync_state_del)) +
1200 sizeof(struct pfsync_header);
1202 case PFSYNC_ACT_UREQ:
1203 len = (sc->sc_maxcount * sizeof(struct pfsync_state_upd_req)) +
1204 sizeof(struct pfsync_header);
1206 case PFSYNC_ACT_BUS:
1207 len = sizeof(struct pfsync_header) +
1208 sizeof(struct pfsync_state_bus);
1210 case PFSYNC_ACT_TDB_UPD:
1211 len = (sc->sc_maxcount * sizeof(struct pfsync_tdb)) +
1212 sizeof(struct pfsync_header);
1215 len = (sc->sc_maxcount * sizeof(struct pfsync_state)) +
1216 sizeof(struct pfsync_header);
1221 MCLGET(m, M_WAITOK);
1222 if ((m->m_flags & M_EXT) == 0) {
1224 sc->sc_if.if_oerrors++;
1227 m->m_data += (MCLBYTES - len) &~ (sizeof(long) - 1);
1231 m->m_pkthdr.rcvif = NULL;
1232 m->m_pkthdr.len = m->m_len = sizeof(struct pfsync_header);
1233 h = mtod(m, struct pfsync_header *);
1234 h->version = PFSYNC_VERSION;
1239 *sp = (void *)((char *)h + PFSYNC_HDRLEN);
1240 lwkt_reltoken(&pf_token);
1241 callout_reset(&sc->sc_tmo, hz, pfsync_timeout,
1242 LIST_FIRST(&pfsync_list));
1243 lwkt_gettoken(&pf_token);
1248 pfsync_pack_state(u_int8_t action, struct pf_state *st, int flags)
1250 struct ifnet *ifp = NULL;
1251 struct pfsync_softc *sc = pfsyncif;
1252 struct pfsync_header *h, *h_net;
1253 struct pfsync_state *sp = NULL;
1254 struct pfsync_state_upd *up = NULL;
1255 struct pfsync_state_del *dp = NULL;
1257 u_int8_t i = 255, newaction = 0;
1264 * If a packet falls in the forest and there's nobody around to
1265 * hear, does it make a sound?
1267 if (ifp->if_bpf == NULL && sc->sc_sync_ifp == NULL &&
1268 sc->sc_sync_peer.s_addr == INADDR_PFSYNC_GROUP) {
1269 /* Don't leave any stale pfsync packets hanging around. */
1270 if (sc->sc_mbuf != NULL) {
1271 m_freem(sc->sc_mbuf);
1273 sc->sc_statep.s = NULL;
1278 if (action >= PFSYNC_ACT_MAX)
1282 if (sc->sc_mbuf == NULL) {
1283 if ((sc->sc_mbuf = pfsync_get_mbuf(sc, action,
1284 (void *)&sc->sc_statep.s)) == NULL) {
1288 h = mtod(sc->sc_mbuf, struct pfsync_header *);
1290 h = mtod(sc->sc_mbuf, struct pfsync_header *);
1291 if (h->action != action) {
1293 if ((sc->sc_mbuf = pfsync_get_mbuf(sc, action,
1294 (void *)&sc->sc_statep.s)) == NULL) {
1298 h = mtod(sc->sc_mbuf, struct pfsync_header *);
1301 * If it's an update, look in the packet to see if
1302 * we already have an update for the state.
1304 if (action == PFSYNC_ACT_UPD && sc->sc_maxupdates) {
1305 struct pfsync_state *usp =
1306 (void *)((char *)h + PFSYNC_HDRLEN);
1308 for (i = 0; i < h->count; i++) {
1309 if (!memcmp(usp->id, &st->id,
1311 usp->creatorid == st->creatorid) {
1322 st->pfsync_time = mycpu->gd_time_seconds;;
1325 /* not a "duplicate" update */
1327 sp = sc->sc_statep.s++;
1328 sc->sc_mbuf->m_pkthdr.len =
1329 sc->sc_mbuf->m_len += sizeof(struct pfsync_state);
1331 bzero(sp, sizeof(*sp));
1333 pfsync_state_export(sp, st);
1335 if (flags & PFSYNC_FLAG_STALE)
1336 sp->sync_flags |= PFSTATE_STALE;
1338 pf_state_peer_hton(&st->src, &sp->src);
1339 pf_state_peer_hton(&st->dst, &sp->dst);
1341 if (st->expire <= time_second)
1342 sp->expire = htonl(0);
1344 sp->expire = htonl(st->expire - time_second);
1347 /* do we need to build "compressed" actions for network transfer? */
1348 if (sc->sc_sync_ifp && flags & PFSYNC_FLAG_COMPRESS) {
1350 case PFSYNC_ACT_UPD:
1351 newaction = PFSYNC_ACT_UPD_C;
1353 case PFSYNC_ACT_DEL:
1354 newaction = PFSYNC_ACT_DEL_C;
1357 /* by default we just send the uncompressed states */
1363 if (sc->sc_mbuf_net == NULL) {
1364 if ((sc->sc_mbuf_net = pfsync_get_mbuf(sc, newaction,
1365 (void *)&sc->sc_statep_net.s)) == NULL) {
1370 h_net = mtod(sc->sc_mbuf_net, struct pfsync_header *);
1372 switch (newaction) {
1373 case PFSYNC_ACT_UPD_C:
1375 up = (void *)((char *)h_net +
1376 PFSYNC_HDRLEN + (i * sizeof(*up)));
1380 sc->sc_mbuf_net->m_pkthdr.len =
1381 sc->sc_mbuf_net->m_len += sizeof(*up);
1382 up = sc->sc_statep_net.u++;
1384 bzero(up, sizeof(*up));
1385 bcopy(&st->id, up->id, sizeof(up->id));
1386 up->creatorid = st->creatorid;
1388 up->timeout = st->timeout;
1389 up->expire = sp->expire;
1393 case PFSYNC_ACT_DEL_C:
1394 sc->sc_mbuf_net->m_pkthdr.len =
1395 sc->sc_mbuf_net->m_len += sizeof(*dp);
1396 dp = sc->sc_statep_net.d++;
1399 bzero(dp, sizeof(*dp));
1400 bcopy(&st->id, dp->id, sizeof(dp->id));
1401 dp->creatorid = st->creatorid;
1406 if (h->count == sc->sc_maxcount ||
1407 (sc->sc_maxupdates && (sp->updates >= sc->sc_maxupdates)))
1408 ret = pfsync_sendout(sc);
1415 pfsync_request_update(struct pfsync_state_upd *up, struct in_addr *src)
1417 struct ifnet *ifp = NULL;
1418 struct pfsync_header *h;
1419 struct pfsync_softc *sc = pfsyncif;
1420 struct pfsync_state_upd_req *rup;
1427 if (sc->sc_mbuf == NULL) {
1428 if ((sc->sc_mbuf = pfsync_get_mbuf(sc, PFSYNC_ACT_UREQ,
1429 (void *)&sc->sc_statep.s)) == NULL)
1431 h = mtod(sc->sc_mbuf, struct pfsync_header *);
1433 h = mtod(sc->sc_mbuf, struct pfsync_header *);
1434 if (h->action != PFSYNC_ACT_UREQ) {
1436 if ((sc->sc_mbuf = pfsync_get_mbuf(sc, PFSYNC_ACT_UREQ,
1437 (void *)&sc->sc_statep.s)) == NULL)
1439 h = mtod(sc->sc_mbuf, struct pfsync_header *);
1444 sc->sc_sendaddr = *src;
1445 sc->sc_mbuf->m_pkthdr.len = sc->sc_mbuf->m_len += sizeof(*rup);
1447 rup = sc->sc_statep.r++;
1448 bzero(rup, sizeof(*rup));
1450 bcopy(up->id, rup->id, sizeof(rup->id));
1451 rup->creatorid = up->creatorid;
1454 if (h->count == sc->sc_maxcount)
1455 ret = pfsync_sendout(sc);
1461 pfsync_clear_states(u_int32_t creatorid, char *ifname)
1463 struct ifnet *ifp = NULL;
1464 struct pfsync_softc *sc = pfsyncif;
1465 struct pfsync_state_clr *cp;
1473 if (sc->sc_mbuf != NULL)
1475 if ((sc->sc_mbuf = pfsync_get_mbuf(sc, PFSYNC_ACT_CLR,
1476 (void *)&sc->sc_statep.c)) == NULL) {
1480 sc->sc_mbuf->m_pkthdr.len = sc->sc_mbuf->m_len += sizeof(*cp);
1481 cp = sc->sc_statep.c;
1482 cp->creatorid = creatorid;
1484 strlcpy(cp->ifname, ifname, IFNAMSIZ);
1486 ret = (pfsync_sendout(sc));
1492 pfsync_timeout(void *v)
1494 struct pfsync_softc *sc = v;
1502 pfsync_send_bus(struct pfsync_softc *sc, u_int8_t status)
1504 struct pfsync_state_bus *bus;
1506 if (sc->sc_mbuf != NULL)
1509 if (pfsync_sync_ok &&
1510 (sc->sc_mbuf = pfsync_get_mbuf(sc, PFSYNC_ACT_BUS,
1511 (void *)&sc->sc_statep.b)) != NULL) {
1512 sc->sc_mbuf->m_pkthdr.len = sc->sc_mbuf->m_len += sizeof(*bus);
1513 bus = sc->sc_statep.b;
1514 bus->creatorid = pf_status.hostid;
1515 bus->status = status;
1516 bus->endtime = htonl(mycpu->gd_time_seconds - sc->sc_ureq_received);
1522 pfsync_bulk_update(void *v)
1524 struct pfsync_softc *sc = v;
1526 struct pf_state *state;
1528 ASSERT_LWKT_TOKEN_HELD(&pf_token);
1531 if (sc->sc_mbuf != NULL)
1535 * Grab at most PFSYNC_BULKPACKETS worth of states which have not
1536 * been sent since the latest request was made.
1538 state = sc->sc_bulk_send_next;
1541 /* send state update if syncable and not already sent */
1542 if (!state->sync_flags
1543 && state->timeout < PFTM_MAX
1544 && state->pfsync_time <= sc->sc_ureq_received) {
1545 pfsync_pack_state(PFSYNC_ACT_UPD, state, 0);
1549 /* figure next state to send */
1550 state = TAILQ_NEXT(state, entry_list);
1552 /* wrap to start of list if we hit the end */
1554 state = TAILQ_FIRST(&state_list);
1555 } while (i < sc->sc_maxcount * PFSYNC_BULKPACKETS &&
1556 state != sc->sc_bulk_terminator);
1558 if (!state || state == sc->sc_bulk_terminator) {
1560 pfsync_send_bus(sc, PFSYNC_BUS_END);
1561 sc->sc_ureq_received = 0;
1562 sc->sc_bulk_send_next = NULL;
1563 sc->sc_bulk_terminator = NULL;
1564 lwkt_reltoken(&pf_token);
1565 callout_stop(&sc->sc_bulk_tmo);
1566 lwkt_gettoken(&pf_token);
1567 if (pf_status.debug >= PF_DEBUG_MISC)
1568 kprintf("pfsync: bulk update complete\n");
1570 /* look again for more in a bit */
1571 lwkt_reltoken(&pf_token);
1572 callout_reset(&sc->sc_bulk_tmo, 1, pfsync_timeout,
1573 LIST_FIRST(&pfsync_list));
1574 lwkt_gettoken(&pf_token);
1575 sc->sc_bulk_send_next = state;
1577 if (sc->sc_mbuf != NULL)
1583 pfsync_bulkfail(void *v)
1585 struct pfsync_softc *sc = v;
1588 ASSERT_LWKT_TOKEN_HELD(&pf_token);
1590 if (sc->sc_bulk_tries++ < PFSYNC_MAX_BULKTRIES) {
1591 /* Try again in a bit */
1592 lwkt_reltoken(&pf_token);
1593 callout_reset(&sc->sc_bulkfail_tmo, 5 * hz, pfsync_bulkfail,
1594 LIST_FIRST(&pfsync_list));
1595 lwkt_gettoken(&pf_token);
1597 error = pfsync_request_update(NULL, NULL);
1598 if (error == ENOMEM) {
1599 if (pf_status.debug >= PF_DEBUG_MISC)
1600 kprintf("pfsync: cannot allocate mbufs for "
1606 /* Pretend like the transfer was ok */
1607 sc->sc_ureq_sent = 0;
1608 sc->sc_bulk_tries = 0;
1610 if (!pfsync_sync_ok)
1611 carp_group_demote_adj(&sc->sc_if, -1);
1614 if (pf_status.debug >= PF_DEBUG_MISC)
1615 kprintf("pfsync: failed to receive "
1616 "bulk update status\n");
1617 lwkt_reltoken(&pf_token);
1618 callout_stop(&sc->sc_bulkfail_tmo);
1619 lwkt_gettoken(&pf_token);
1623 /* This must be called in splnet() */
1625 pfsync_sendout(struct pfsync_softc *sc)
1628 struct ifnet *ifp = &sc->sc_if;
1632 ASSERT_LWKT_TOKEN_HELD(&pf_token);
1634 lwkt_reltoken(&pf_token);
1635 callout_stop(&sc->sc_tmo);
1636 lwkt_gettoken(&pf_token);
1638 if (sc->sc_mbuf == NULL)
1642 sc->sc_statep.s = NULL;
1646 lwkt_reltoken(&pf_token);
1647 bpf_mtap(ifp->if_bpf, m, BPF_DIRECTION_OUT);
1648 lwkt_gettoken(&pf_token);
1652 if (sc->sc_mbuf_net) {
1654 m = sc->sc_mbuf_net;
1655 sc->sc_mbuf_net = NULL;
1656 sc->sc_statep_net.s = NULL;
1659 return pfsync_sendout_mbuf(sc, m);
1663 pfsync_sendout_mbuf(struct pfsync_softc *sc, struct mbuf *m)
1668 if (sc->sc_sync_ifp ||
1669 sc->sc_sync_peer.s_addr != INADDR_PFSYNC_GROUP) {
1670 M_PREPEND(m, sizeof(struct ip), M_WAITOK);
1672 pfsyncstats.pfsyncs_onomem++;
1675 ip = mtod(m, struct ip *);
1676 ip->ip_v = IPVERSION;
1677 ip->ip_hl = sizeof(*ip) >> 2;
1678 ip->ip_tos = IPTOS_LOWDELAY;
1679 ip->ip_len = htons(m->m_pkthdr.len);
1680 ip->ip_id = htons(ip_randomid());
1681 ip->ip_off = htons(IP_DF);
1682 ip->ip_ttl = PFSYNC_DFLTTL;
1683 ip->ip_p = IPPROTO_PFSYNC;
1686 bzero(&sa, sizeof(sa));
1687 ip->ip_src.s_addr = INADDR_ANY;
1689 if (sc->sc_sendaddr.s_addr == INADDR_PFSYNC_GROUP)
1690 m->m_flags |= M_MCAST;
1691 ip->ip_dst = sc->sc_sendaddr;
1692 sc->sc_sendaddr.s_addr = sc->sc_sync_peer.s_addr;
1694 pfsyncstats.pfsyncs_opackets++;
1696 if (ip_output(m, NULL, NULL, IP_RAWOUTPUT, &sc->sc_imo, NULL))
1697 pfsyncstats.pfsyncs_oerrors++;
1705 pfsync_modevent(module_t mod, int type, void *data)
1709 struct pfsync_softc *pfs_if, *tmp;
1711 lwkt_gettoken(&pf_token);
1715 LIST_INIT(&pfsync_list);
1716 lwkt_reltoken(&pf_token);
1717 if_clone_attach(&pfsync_cloner);
1718 lwkt_gettoken(&pf_token);
1719 /* Override the function pointer for pf_ioctl.c */
1723 lwkt_reltoken(&pf_token);
1724 if_clone_detach(&pfsync_cloner);
1725 lwkt_gettoken(&pf_token);
1726 LIST_FOREACH_MUTABLE(pfs_if, &pfsync_list, sc_next, tmp) {
1727 pfsync_clone_destroy(&pfs_if->sc_if);
1736 lwkt_reltoken(&pf_token);
1740 static moduledata_t pfsync_mod = {
1746 #define PFSYNC_MODVER 44
1748 DECLARE_MODULE(pfsync, pfsync_mod, SI_SUB_PSEUDO, SI_ORDER_ANY);
1749 MODULE_VERSION(pfsync, PFSYNC_MODVER);