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/ifq_var.h>
52 #include <net/route.h>
54 #include <netinet/in.h>
55 #include <netinet/if_ether.h>
56 #include <netinet/ip_carp.h>
57 #include <netinet/tcp.h>
58 #include <netinet/tcp_seq.h>
61 #include <netinet/in_systm.h>
62 #include <netinet/in_var.h>
63 #include <netinet/ip.h>
64 #include <netinet/ip_var.h>
68 #include <netinet6/nd6.h>
71 #include <net/pf/pfvar.h>
72 #include <net/pf/if_pfsync.h>
74 #define PFSYNCNAME "pfsync"
76 #define PFSYNC_MINMTU \
77 (sizeof(struct pfsync_header) + sizeof(struct pf_state))
80 #define DPRINTF(x) do { if (pfsyncdebug) kprintf x ; } while (0)
86 struct pfsync_softc *pfsyncif = NULL;
87 struct pfsyncstats pfsyncstats;
89 void pfsyncattach(int);
90 static int pfsync_clone_destroy(struct ifnet *);
91 static int pfsync_clone_create(struct if_clone *, int, caddr_t);
92 void pfsync_setmtu(struct pfsync_softc *, int);
93 int pfsync_alloc_scrub_memory(struct pfsync_state_peer *,
94 struct pf_state_peer *);
95 int pfsyncoutput(struct ifnet *, struct mbuf *, struct sockaddr *,
97 int pfsyncioctl(struct ifnet *, u_long, caddr_t, struct ucred *);
98 void pfsyncstart(struct ifnet *, struct ifaltq_subque *);
100 struct mbuf *pfsync_get_mbuf(struct pfsync_softc *, u_int8_t, void **);
101 int pfsync_request_update(struct pfsync_state_upd *, struct in_addr *);
102 int pfsync_sendout(struct pfsync_softc *);
103 int pfsync_sendout_mbuf(struct pfsync_softc *, struct mbuf *);
104 void pfsync_timeout(void *);
105 void pfsync_send_bus(struct pfsync_softc *, u_int8_t);
106 void pfsync_bulk_update(void *);
107 void pfsync_bulkfail(void *);
109 static MALLOC_DEFINE(M_PFSYNC, PFSYNCNAME, "Packet Filter State Sync. Interface");
110 static LIST_HEAD(pfsync_list, pfsync_softc) pfsync_list;
114 struct if_clone pfsync_cloner =
115 IF_CLONE_INITIALIZER("pfsync", pfsync_clone_create, pfsync_clone_destroy, 1 ,1);
118 pfsyncattach(int npfsync)
120 if_clone_attach(&pfsync_cloner);
123 pfsync_clone_create(struct if_clone *ifc, int unit, caddr_t param __unused)
125 struct pfsync_softc *sc;
128 lwkt_gettoken(&pf_token);
130 sc = kmalloc(sizeof(*sc), M_PFSYNC, M_WAITOK | M_ZERO);
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 ifq_set_maxlen(&ifp->if_snd, 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, struct ifaltq_subque *ifsq)
215 ASSERT_ALTQ_SQ_DEFAULT(ifp, ifsq);
220 pfsync_alloc_scrub_memory(struct pfsync_state_peer *s,
221 struct pf_state_peer *d)
223 if (s->scrub.scrub_flag && d->scrub == NULL) {
224 d->scrub = kmalloc(sizeof(struct pf_state_scrub), M_PFSYNC, M_NOWAIT|M_ZERO);
226 if (d->scrub == NULL)
234 pfsync_state_export(struct pfsync_state *sp, struct pf_state *st)
236 bzero(sp, sizeof(struct pfsync_state));
238 /* copy from state key */
239 sp->key[PF_SK_WIRE].addr[0] = st->key[PF_SK_WIRE]->addr[0];
240 sp->key[PF_SK_WIRE].addr[1] = st->key[PF_SK_WIRE]->addr[1];
241 sp->key[PF_SK_WIRE].port[0] = st->key[PF_SK_WIRE]->port[0];
242 sp->key[PF_SK_WIRE].port[1] = st->key[PF_SK_WIRE]->port[1];
243 sp->key[PF_SK_STACK].addr[0] = st->key[PF_SK_STACK]->addr[0];
244 sp->key[PF_SK_STACK].addr[1] = st->key[PF_SK_STACK]->addr[1];
245 sp->key[PF_SK_STACK].port[0] = st->key[PF_SK_STACK]->port[0];
246 sp->key[PF_SK_STACK].port[1] = st->key[PF_SK_STACK]->port[1];
247 sp->proto = st->key[PF_SK_WIRE]->proto;
248 sp->af = st->key[PF_SK_WIRE]->af;
250 /* copy from state */
251 strlcpy(sp->ifname, st->kif->pfik_name, sizeof(sp->ifname));
252 bcopy(&st->rt_addr, &sp->rt_addr, sizeof(sp->rt_addr));
253 sp->creation = htonl(time_second - st->creation);
254 sp->expire = pf_state_expires(st);
255 if (sp->expire <= time_second)
256 sp->expire = htonl(0);
258 sp->expire = htonl(sp->expire - time_second);
260 sp->direction = st->direction;
262 sp->timeout = st->timeout;
263 sp->state_flags = st->state_flags;
265 sp->sync_flags |= PFSYNC_FLAG_SRCNODE;
266 if (st->nat_src_node)
267 sp->sync_flags |= PFSYNC_FLAG_NATSRCNODE;
269 bcopy(&st->id, &sp->id, sizeof(sp->id));
270 sp->creatorid = st->creatorid;
271 pf_state_peer_hton(&st->src, &sp->src);
272 pf_state_peer_hton(&st->dst, &sp->dst);
274 if (st->rule.ptr == NULL)
275 sp->rule = htonl(-1);
277 sp->rule = htonl(st->rule.ptr->nr);
278 if (st->anchor.ptr == NULL)
279 sp->anchor = htonl(-1);
281 sp->anchor = htonl(st->anchor.ptr->nr);
282 if (st->nat_rule.ptr == NULL)
283 sp->nat_rule = htonl(-1);
285 sp->nat_rule = htonl(st->nat_rule.ptr->nr);
287 pf_state_counter_hton(st->packets[0], sp->packets[0]);
288 pf_state_counter_hton(st->packets[1], sp->packets[1]);
289 pf_state_counter_hton(st->bytes[0], sp->bytes[0]);
290 pf_state_counter_hton(st->bytes[1], sp->bytes[1]);
295 pfsync_state_import(struct pfsync_state *sp, u_int8_t flags)
297 struct pf_state *st = NULL;
298 struct pf_state_key *skw = NULL, *sks = NULL;
299 struct pf_rule *r = NULL;
304 if (sp->creatorid == 0 && pf_status.debug >= PF_DEBUG_MISC) {
305 kprintf("pfsync_insert_net_state: invalid creator id:"
306 " %08x\n", ntohl(sp->creatorid));
310 if ((kif = pfi_kif_get(sp->ifname)) == NULL) {
311 if (pf_status.debug >= PF_DEBUG_MISC)
312 kprintf("pfsync_insert_net_state: "
313 "unknown interface: %s\n", sp->ifname);
314 if (flags & PFSYNC_SI_IOCTL)
316 return (0); /* skip this state */
320 * If the ruleset checksums match or the state is coming from the ioctl,
321 * it's safe to associate the state with the rule of that number.
323 if (sp->rule != htonl(-1) && sp->anchor == htonl(-1) &&
324 (flags & (PFSYNC_SI_IOCTL | PFSYNC_SI_CKSUM)) && ntohl(sp->rule) <
325 pf_main_ruleset.rules[PF_RULESET_FILTER].active.rcount)
326 r = pf_main_ruleset.rules[
327 PF_RULESET_FILTER].active.ptr_array[ntohl(sp->rule)];
329 r = &pf_default_rule;
331 if ((r->max_states && r->states_cur >= r->max_states))
334 if (flags & PFSYNC_SI_IOCTL)
335 pool_flags = M_WAITOK | M_NULLOK | M_ZERO;
337 pool_flags = M_WAITOK | M_ZERO;
339 if ((st = kmalloc(sizeof(struct pf_state), M_PFSYNC, pool_flags)) == NULL)
342 if ((skw = pf_alloc_state_key(pool_flags)) == NULL)
345 if (PF_ANEQ(&sp->key[PF_SK_WIRE].addr[0],
346 &sp->key[PF_SK_STACK].addr[0], sp->af) ||
347 PF_ANEQ(&sp->key[PF_SK_WIRE].addr[1],
348 &sp->key[PF_SK_STACK].addr[1], sp->af) ||
349 sp->key[PF_SK_WIRE].port[0] != sp->key[PF_SK_STACK].port[0] ||
350 sp->key[PF_SK_WIRE].port[1] != sp->key[PF_SK_STACK].port[1]) {
351 if ((sks = pf_alloc_state_key(pool_flags)) == NULL)
356 /* allocate memory for scrub info */
357 if (pfsync_alloc_scrub_memory(&sp->src, &st->src) ||
358 pfsync_alloc_scrub_memory(&sp->dst, &st->dst))
361 /* copy to state key(s) */
362 skw->addr[0] = sp->key[PF_SK_WIRE].addr[0];
363 skw->addr[1] = sp->key[PF_SK_WIRE].addr[1];
364 skw->port[0] = sp->key[PF_SK_WIRE].port[0];
365 skw->port[1] = sp->key[PF_SK_WIRE].port[1];
366 skw->proto = sp->proto;
369 sks->addr[0] = sp->key[PF_SK_STACK].addr[0];
370 sks->addr[1] = sp->key[PF_SK_STACK].addr[1];
371 sks->port[0] = sp->key[PF_SK_STACK].port[0];
372 sks->port[1] = sp->key[PF_SK_STACK].port[1];
373 sks->proto = sp->proto;
378 bcopy(&sp->rt_addr, &st->rt_addr, sizeof(st->rt_addr));
379 st->creation = time_second - ntohl(sp->creation);
380 st->expire = time_second;
382 /* XXX No adaptive scaling. */
383 st->expire -= r->timeout[sp->timeout] - ntohl(sp->expire);
386 st->expire = ntohl(sp->expire) + time_second;
387 st->direction = sp->direction;
389 st->timeout = sp->timeout;
390 st->state_flags = sp->state_flags;
391 if (!(flags & PFSYNC_SI_IOCTL))
392 st->sync_flags = PFSTATE_FROMSYNC;
394 bcopy(sp->id, &st->id, sizeof(st->id));
395 st->creatorid = sp->creatorid;
396 pf_state_peer_ntoh(&sp->src, &st->src);
397 pf_state_peer_ntoh(&sp->dst, &st->dst);
400 st->nat_rule.ptr = NULL;
401 st->anchor.ptr = NULL;
407 /* XXX when we have nat_rule/anchors, use STATE_INC_COUNTERS */
411 if ((error = pf_state_insert(kif, skw, sks, st)) != 0) {
412 /* XXX when we have nat_rule/anchors, use STATE_DEC_COUNTERS */
424 kfree(skw, M_PFSYNC);
426 kfree(sks, M_PFSYNC);
428 cleanup_state: /* pf_state_insert frees the state keys */
431 kfree(st->dst.scrub, M_PFSYNC);
433 kfree(st->src.scrub, M_PFSYNC);
440 pfsync_input(struct mbuf *m, ...)
442 struct ip *ip = mtod(m, struct ip *);
443 struct pfsync_header *ph;
444 struct pfsync_softc *sc = pfsyncif;
446 struct pf_state_key *sk;
447 struct pf_state_item *si;
448 struct pf_state_cmp id_key;
449 struct pfsync_state *sp;
450 struct pfsync_state_upd *up;
451 struct pfsync_state_del *dp;
452 struct pfsync_state_clr *cp;
453 struct pfsync_state_upd_req *rup;
454 struct pfsync_state_bus *bus;
456 struct pfsync_tdb *pt;
460 int iplen, action, error, i, count, offp, sfail, stale = 0;
463 /* This function is not yet called from anywhere */
464 /* Still we assume for safety that pf_token must be held */
465 ASSERT_LWKT_TOKEN_HELD(&pf_token);
467 pfsyncstats.pfsyncs_ipackets++;
469 /* verify that we have a sync interface configured */
470 if (!sc || !sc->sc_sync_ifp || !pf_status.running)
473 /* verify that the packet came in on the right interface */
474 if (sc->sc_sync_ifp != m->m_pkthdr.rcvif) {
475 pfsyncstats.pfsyncs_badif++;
479 /* verify that the IP TTL is 255. */
480 if (ip->ip_ttl != PFSYNC_DFLTTL) {
481 pfsyncstats.pfsyncs_badttl++;
485 iplen = ip->ip_hl << 2;
487 if (m->m_pkthdr.len < iplen + sizeof(*ph)) {
488 pfsyncstats.pfsyncs_hdrops++;
492 if (iplen + sizeof(*ph) > m->m_len) {
493 if ((m = m_pullup(m, iplen + sizeof(*ph))) == NULL) {
494 pfsyncstats.pfsyncs_hdrops++;
497 ip = mtod(m, struct ip *);
499 ph = (struct pfsync_header *)((char *)ip + iplen);
501 /* verify the version */
502 if (ph->version != PFSYNC_VERSION) {
503 pfsyncstats.pfsyncs_badver++;
510 /* make sure it's a valid action code */
511 if (action >= PFSYNC_ACT_MAX) {
512 pfsyncstats.pfsyncs_badact++;
516 /* Cheaper to grab this now than having to mess with mbufs later */
519 if (!bcmp(&ph->pf_chksum, &pf_status.pf_chksum, PF_MD5_DIGEST_LENGTH))
520 flags |= PFSYNC_SI_CKSUM;
523 case PFSYNC_ACT_CLR: {
524 struct pf_state *nexts;
525 struct pf_state_key *nextsk;
528 if ((mp = m_pulldown(m, iplen + sizeof(*ph),
529 sizeof(*cp), &offp)) == NULL) {
530 pfsyncstats.pfsyncs_badlen++;
533 cp = (struct pfsync_state_clr *)(mp->m_data + offp);
534 creatorid = cp->creatorid;
537 if (cp->ifname[0] == '\0') {
538 for (st = RB_MIN(pf_state_tree_id, &tree_id);
540 nexts = RB_NEXT(pf_state_tree_id, &tree_id, st);
541 if (st->creatorid == creatorid) {
542 st->sync_flags |= PFSTATE_FROMSYNC;
547 if ((kif = pfi_kif_get(cp->ifname)) == NULL) {
552 for (sk = RB_MIN(pf_state_tree,
553 &pf_statetbl); sk; sk = nextsk) {
554 nextsk = RB_NEXT(pf_state_tree,
556 TAILQ_FOREACH(si, &sk->states, entry) {
557 if (si->s->creatorid == creatorid) {
560 pf_unlink_state(si->s);
570 if ((mp = m_pulldown(m, iplen + sizeof(*ph),
571 count * sizeof(*sp), &offp)) == NULL) {
572 pfsyncstats.pfsyncs_badlen++;
577 for (i = 0, sp = (struct pfsync_state *)(mp->m_data + offp);
578 i < count; i++, sp++) {
579 /* check for invalid values */
580 if (sp->timeout >= PFTM_MAX ||
581 sp->src.state > PF_TCPS_PROXY_DST ||
582 sp->dst.state > PF_TCPS_PROXY_DST ||
583 sp->direction > PF_OUT ||
584 (sp->af != AF_INET && sp->af != AF_INET6)) {
585 if (pf_status.debug >= PF_DEBUG_MISC)
586 kprintf("pfsync_insert: PFSYNC_ACT_INS: "
588 pfsyncstats.pfsyncs_badval++;
592 if ((error = pfsync_state_import(sp, flags))) {
593 if (error == ENOMEM) {
603 if ((mp = m_pulldown(m, iplen + sizeof(*ph),
604 count * sizeof(*sp), &offp)) == NULL) {
605 pfsyncstats.pfsyncs_badlen++;
610 for (i = 0, sp = (struct pfsync_state *)(mp->m_data + offp);
611 i < count; i++, sp++) {
612 int flags = PFSYNC_FLAG_STALE;
614 /* check for invalid values */
615 if (sp->timeout >= PFTM_MAX ||
616 sp->src.state > PF_TCPS_PROXY_DST ||
617 sp->dst.state > PF_TCPS_PROXY_DST) {
618 if (pf_status.debug >= PF_DEBUG_MISC)
619 kprintf("pfsync_insert: PFSYNC_ACT_UPD: "
621 pfsyncstats.pfsyncs_badval++;
625 bcopy(sp->id, &id_key.id, sizeof(id_key.id));
626 id_key.creatorid = sp->creatorid;
628 st = pf_find_state_byid(&id_key);
630 /* insert the update */
631 if (pfsync_state_import(sp, flags))
632 pfsyncstats.pfsyncs_badstate++;
635 sk = st->key[PF_SK_WIRE]; /* XXX right one? */
637 if (sk->proto == IPPROTO_TCP) {
639 * The state should never go backwards except
640 * for syn-proxy states. Neither should the
641 * sequence window slide backwards.
643 if (st->src.state > sp->src.state &&
644 (st->src.state < PF_TCPS_PROXY_SRC ||
645 sp->src.state >= PF_TCPS_PROXY_SRC))
647 else if (SEQ_GT(st->src.seqlo,
648 ntohl(sp->src.seqlo)))
650 else if (st->dst.state > sp->dst.state) {
651 /* There might still be useful
652 * information about the src state here,
653 * so import that part of the update,
654 * then "fail" so we send the updated
655 * state back to the peer who is missing
656 * our what we know. */
657 pf_state_peer_ntoh(&sp->src, &st->src);
658 /* XXX do anything with timeouts? */
661 } else if (st->dst.state >= TCPS_SYN_SENT &&
662 SEQ_GT(st->dst.seqlo, ntohl(sp->dst.seqlo)))
666 * Non-TCP protocol state machine always go
669 if (st->src.state > sp->src.state)
671 else if (st->dst.state > sp->dst.state)
675 if (pf_status.debug >= PF_DEBUG_MISC)
676 kprintf("pfsync: %s stale update "
679 (sfail < 7 ? "ignoring"
681 (uintmax_t)be64toh(st->id),
682 ntohl(st->creatorid));
683 pfsyncstats.pfsyncs_stale++;
685 if (!(sp->sync_flags & PFSTATE_STALE)) {
686 /* we have a better state, send it */
687 if (sc->sc_mbuf != NULL && !stale)
692 PFSYNC_ACT_UPD, st, flags);
696 pfsync_alloc_scrub_memory(&sp->dst, &st->dst);
697 pf_state_peer_ntoh(&sp->src, &st->src);
698 pf_state_peer_ntoh(&sp->dst, &st->dst);
699 st->expire = ntohl(sp->expire) + time_second;
700 st->timeout = sp->timeout;
702 if (stale && sc->sc_mbuf != NULL)
707 * It's not strictly necessary for us to support the "uncompressed"
708 * delete action, but it's relatively simple and maintains consistency.
711 if ((mp = m_pulldown(m, iplen + sizeof(*ph),
712 count * sizeof(*sp), &offp)) == NULL) {
713 pfsyncstats.pfsyncs_badlen++;
718 for (i = 0, sp = (struct pfsync_state *)(mp->m_data + offp);
719 i < count; i++, sp++) {
720 bcopy(sp->id, &id_key.id, sizeof(id_key.id));
721 id_key.creatorid = sp->creatorid;
723 st = pf_find_state_byid(&id_key);
725 pfsyncstats.pfsyncs_badstate++;
728 st->sync_flags |= PFSTATE_FROMSYNC;
733 case PFSYNC_ACT_UPD_C: {
734 int update_requested = 0;
736 if ((mp = m_pulldown(m, iplen + sizeof(*ph),
737 count * sizeof(*up), &offp)) == NULL) {
738 pfsyncstats.pfsyncs_badlen++;
743 for (i = 0, up = (struct pfsync_state_upd *)(mp->m_data + offp);
744 i < count; i++, up++) {
745 /* check for invalid values */
746 if (up->timeout >= PFTM_MAX ||
747 up->src.state > PF_TCPS_PROXY_DST ||
748 up->dst.state > PF_TCPS_PROXY_DST) {
749 if (pf_status.debug >= PF_DEBUG_MISC)
750 kprintf("pfsync_insert: "
753 pfsyncstats.pfsyncs_badval++;
757 bcopy(up->id, &id_key.id, sizeof(id_key.id));
758 id_key.creatorid = up->creatorid;
760 st = pf_find_state_byid(&id_key);
762 /* We don't have this state. Ask for it. */
763 error = pfsync_request_update(up, &src);
764 if (error == ENOMEM) {
768 update_requested = 1;
769 pfsyncstats.pfsyncs_badstate++;
772 sk = st->key[PF_SK_WIRE]; /* XXX right one? */
774 if (sk->proto == IPPROTO_TCP) {
776 * The state should never go backwards except
777 * for syn-proxy states. Neither should the
778 * sequence window slide backwards.
780 if (st->src.state > up->src.state &&
781 (st->src.state < PF_TCPS_PROXY_SRC ||
782 up->src.state >= PF_TCPS_PROXY_SRC))
784 else if (st->dst.state > up->dst.state)
786 else if (SEQ_GT(st->src.seqlo,
787 ntohl(up->src.seqlo)))
789 else if (st->dst.state >= TCPS_SYN_SENT &&
790 SEQ_GT(st->dst.seqlo, ntohl(up->dst.seqlo)))
794 * Non-TCP protocol state machine always go
797 if (st->src.state > up->src.state)
799 else if (st->dst.state > up->dst.state)
803 if (pf_status.debug >= PF_DEBUG_MISC)
804 kprintf("pfsync: ignoring stale update "
805 "(%d) id: %016" PRIx64 " "
806 "creatorid: %08x\n", sfail,
808 ntohl(st->creatorid));
809 pfsyncstats.pfsyncs_stale++;
811 /* we have a better state, send it out */
812 if ((!stale || update_requested) &&
813 sc->sc_mbuf != NULL) {
815 update_requested = 0;
819 pfsync_pack_state(PFSYNC_ACT_UPD, st,
823 pfsync_alloc_scrub_memory(&up->dst, &st->dst);
824 pf_state_peer_ntoh(&up->src, &st->src);
825 pf_state_peer_ntoh(&up->dst, &st->dst);
826 st->expire = ntohl(up->expire) + time_second;
827 st->timeout = up->timeout;
829 if ((update_requested || stale) && sc->sc_mbuf)
834 case PFSYNC_ACT_DEL_C:
835 if ((mp = m_pulldown(m, iplen + sizeof(*ph),
836 count * sizeof(*dp), &offp)) == NULL) {
837 pfsyncstats.pfsyncs_badlen++;
842 for (i = 0, dp = (struct pfsync_state_del *)(mp->m_data + offp);
843 i < count; i++, dp++) {
844 bcopy(dp->id, &id_key.id, sizeof(id_key.id));
845 id_key.creatorid = dp->creatorid;
847 st = pf_find_state_byid(&id_key);
849 pfsyncstats.pfsyncs_badstate++;
852 st->sync_flags |= PFSTATE_FROMSYNC;
857 case PFSYNC_ACT_INS_F:
858 case PFSYNC_ACT_DEL_F:
859 /* not implemented */
861 case PFSYNC_ACT_UREQ:
862 if ((mp = m_pulldown(m, iplen + sizeof(*ph),
863 count * sizeof(*rup), &offp)) == NULL) {
864 pfsyncstats.pfsyncs_badlen++;
869 if (sc->sc_mbuf != NULL)
872 rup = (struct pfsync_state_upd_req *)(mp->m_data + offp);
873 i < count; i++, rup++) {
874 bcopy(rup->id, &id_key.id, sizeof(id_key.id));
875 id_key.creatorid = rup->creatorid;
877 if (id_key.id == 0 && id_key.creatorid == 0) {
878 sc->sc_ureq_received = mycpu->gd_time_seconds;
879 if (sc->sc_bulk_send_next == NULL)
880 sc->sc_bulk_send_next =
881 TAILQ_FIRST(&state_list);
882 sc->sc_bulk_terminator = sc->sc_bulk_send_next;
883 if (pf_status.debug >= PF_DEBUG_MISC)
884 kprintf("pfsync: received "
885 "bulk update request\n");
886 pfsync_send_bus(sc, PFSYNC_BUS_START);
887 lwkt_reltoken(&pf_token);
888 callout_init(&sc->sc_bulk_tmo);
889 lwkt_gettoken(&pf_token);
891 st = pf_find_state_byid(&id_key);
893 pfsyncstats.pfsyncs_badstate++;
897 pfsync_pack_state(PFSYNC_ACT_UPD,
901 if (sc->sc_mbuf != NULL)
906 /* If we're not waiting for a bulk update, who cares. */
907 if (sc->sc_ureq_sent == 0)
910 if ((mp = m_pulldown(m, iplen + sizeof(*ph),
911 sizeof(*bus), &offp)) == NULL) {
912 pfsyncstats.pfsyncs_badlen++;
915 bus = (struct pfsync_state_bus *)(mp->m_data + offp);
916 switch (bus->status) {
917 case PFSYNC_BUS_START:
918 lwkt_reltoken(&pf_token);
919 callout_reset(&sc->sc_bulkfail_tmo,
920 pf_pool_limits[PF_LIMIT_STATES].limit /
921 (PFSYNC_BULKPACKETS * sc->sc_maxcount),
922 pfsync_bulkfail, LIST_FIRST(&pfsync_list));
923 lwkt_gettoken(&pf_token);
924 if (pf_status.debug >= PF_DEBUG_MISC)
925 kprintf("pfsync: received bulk "
929 if (mycpu->gd_time_seconds - ntohl(bus->endtime) >=
931 /* that's it, we're happy */
932 sc->sc_ureq_sent = 0;
933 sc->sc_bulk_tries = 0;
934 lwkt_reltoken(&pf_token);
935 callout_stop(&sc->sc_bulkfail_tmo);
936 lwkt_gettoken(&pf_token);
938 if (!pfsync_sync_ok) {
939 lwkt_reltoken(&pf_token);
940 carp_group_demote_adj(&sc->sc_if, -1);
941 lwkt_gettoken(&pf_token);
945 if (pf_status.debug >= PF_DEBUG_MISC)
946 kprintf("pfsync: received valid "
947 "bulk update end\n");
949 if (pf_status.debug >= PF_DEBUG_MISC)
950 kprintf("pfsync: received invalid "
951 "bulk update end: bad timestamp\n");
957 case PFSYNC_ACT_TDB_UPD:
958 if ((mp = m_pulldown(m, iplen + sizeof(*ph),
959 count * sizeof(*pt), &offp)) == NULL) {
960 pfsyncstats.pfsyncs_badlen++;
964 for (i = 0, pt = (struct pfsync_tdb *)(mp->m_data + offp);
965 i < count; i++, pt++)
966 pfsync_update_net_tdb(pt);
978 pfsyncoutput(struct ifnet *ifp, struct mbuf *m, struct sockaddr *dst,
987 pfsyncioctl(struct ifnet *ifp, u_long cmd, caddr_t data, struct ucred *cr)
989 struct pfsync_softc *sc = ifp->if_softc;
990 struct ifreq *ifr = (struct ifreq *)data;
991 struct ip_moptions *imo = &sc->sc_imo;
992 struct pfsyncreq pfsyncr;
996 lwkt_gettoken(&pf_token);
1001 case SIOCSIFDSTADDR:
1003 if (ifp->if_flags & IFF_UP)
1004 ifp->if_flags |= IFF_RUNNING;
1006 ifp->if_flags &= ~IFF_RUNNING;
1009 if (ifr->ifr_mtu < PFSYNC_MINMTU) {
1010 lwkt_reltoken(&pf_token);
1013 if (ifr->ifr_mtu > MCLBYTES)
1014 ifr->ifr_mtu = MCLBYTES;
1016 if (ifr->ifr_mtu < ifp->if_mtu)
1018 pfsync_setmtu(sc, ifr->ifr_mtu);
1022 bzero(&pfsyncr, sizeof(pfsyncr));
1023 if (sc->sc_sync_ifp)
1024 strlcpy(pfsyncr.pfsyncr_syncdev,
1025 sc->sc_sync_ifp->if_xname, IFNAMSIZ);
1026 pfsyncr.pfsyncr_syncpeer = sc->sc_sync_peer;
1027 pfsyncr.pfsyncr_maxupdates = sc->sc_maxupdates;
1028 lwkt_reltoken(&pf_token);
1029 if ((error = copyout(&pfsyncr, ifr->ifr_data, sizeof(pfsyncr))))
1031 lwkt_gettoken(&pf_token);
1034 if ((error = priv_check_cred(cr, PRIV_ROOT, NULL_CRED_OKAY)) != 0) {
1035 lwkt_reltoken(&pf_token);
1038 if ((error = copyin(ifr->ifr_data, &pfsyncr, sizeof(pfsyncr)))) {
1039 lwkt_reltoken(&pf_token);
1043 if (pfsyncr.pfsyncr_syncpeer.s_addr == 0)
1044 sc->sc_sync_peer.s_addr = INADDR_PFSYNC_GROUP;
1046 sc->sc_sync_peer.s_addr =
1047 pfsyncr.pfsyncr_syncpeer.s_addr;
1049 if (pfsyncr.pfsyncr_maxupdates > 255) {
1050 lwkt_reltoken(&pf_token);
1053 sc->sc_maxupdates = pfsyncr.pfsyncr_maxupdates;
1055 if (pfsyncr.pfsyncr_syncdev[0] == 0) {
1056 sc->sc_sync_ifp = NULL;
1057 if (sc->sc_mbuf_net != NULL) {
1058 /* Don't keep stale pfsync packets around. */
1060 m_freem(sc->sc_mbuf_net);
1061 sc->sc_mbuf_net = NULL;
1062 sc->sc_statep_net.s = NULL;
1065 if (imo->imo_num_memberships > 0) {
1066 in_delmulti(imo->imo_membership[--imo->imo_num_memberships]);
1067 imo->imo_multicast_ifp = NULL;
1072 if ((sifp = ifunit(pfsyncr.pfsyncr_syncdev)) == NULL) {
1073 lwkt_reltoken(&pf_token);
1078 if (sifp->if_mtu < sc->sc_if.if_mtu ||
1079 (sc->sc_sync_ifp != NULL &&
1080 sifp->if_mtu < sc->sc_sync_ifp->if_mtu) ||
1081 sifp->if_mtu < MCLBYTES - sizeof(struct ip))
1083 sc->sc_sync_ifp = sifp;
1085 pfsync_setmtu(sc, sc->sc_if.if_mtu);
1087 if (imo->imo_num_memberships > 0) {
1088 in_delmulti(imo->imo_membership[--imo->imo_num_memberships]);
1089 imo->imo_multicast_ifp = NULL;
1092 if (sc->sc_sync_ifp &&
1093 sc->sc_sync_peer.s_addr == INADDR_PFSYNC_GROUP) {
1094 struct in_addr addr;
1096 if (!(sc->sc_sync_ifp->if_flags & IFF_MULTICAST)) {
1097 sc->sc_sync_ifp = NULL;
1098 lwkt_reltoken(&pf_token);
1100 return (EADDRNOTAVAIL);
1103 addr.s_addr = INADDR_PFSYNC_GROUP;
1105 if ((imo->imo_membership[0] =
1106 in_addmulti(&addr, sc->sc_sync_ifp)) == NULL) {
1107 sc->sc_sync_ifp = NULL;
1108 lwkt_reltoken(&pf_token);
1112 imo->imo_num_memberships++;
1113 imo->imo_multicast_ifp = sc->sc_sync_ifp;
1114 imo->imo_multicast_ttl = PFSYNC_DFLTTL;
1115 imo->imo_multicast_loop = 0;
1118 if (sc->sc_sync_ifp ||
1119 sc->sc_sendaddr.s_addr != INADDR_PFSYNC_GROUP) {
1120 /* Request a full state table update. */
1121 sc->sc_ureq_sent = mycpu->gd_time_seconds;
1124 carp_group_demote_adj(&sc->sc_if, 1);
1127 if (pf_status.debug >= PF_DEBUG_MISC)
1128 kprintf("pfsync: requesting bulk update\n");
1129 lwkt_reltoken(&pf_token);
1130 callout_reset(&sc->sc_bulkfail_tmo, 5 * hz,
1131 pfsync_bulkfail, LIST_FIRST(&pfsync_list));
1132 lwkt_gettoken(&pf_token);
1133 error = pfsync_request_update(NULL, NULL);
1134 if (error == ENOMEM) {
1135 lwkt_reltoken(&pf_token);
1146 lwkt_reltoken(&pf_token);
1150 lwkt_reltoken(&pf_token);
1155 pfsync_setmtu(struct pfsync_softc *sc, int mtu_req)
1159 if (sc->sc_sync_ifp && sc->sc_sync_ifp->if_mtu < mtu_req)
1160 mtu = sc->sc_sync_ifp->if_mtu;
1164 sc->sc_maxcount = (mtu - sizeof(struct pfsync_header)) /
1165 sizeof(struct pfsync_state);
1166 if (sc->sc_maxcount > 254)
1167 sc->sc_maxcount = 254;
1168 sc->sc_if.if_mtu = sizeof(struct pfsync_header) +
1169 sc->sc_maxcount * sizeof(struct pfsync_state);
1173 pfsync_get_mbuf(struct pfsync_softc *sc, u_int8_t action, void **sp)
1175 struct pfsync_header *h;
1179 ASSERT_LWKT_TOKEN_HELD(&pf_token);
1181 MGETHDR(m, M_WAITOK, MT_DATA);
1183 sc->sc_if.if_oerrors++;
1188 case PFSYNC_ACT_CLR:
1189 len = sizeof(struct pfsync_header) +
1190 sizeof(struct pfsync_state_clr);
1192 case PFSYNC_ACT_UPD_C:
1193 len = (sc->sc_maxcount * sizeof(struct pfsync_state_upd)) +
1194 sizeof(struct pfsync_header);
1196 case PFSYNC_ACT_DEL_C:
1197 len = (sc->sc_maxcount * sizeof(struct pfsync_state_del)) +
1198 sizeof(struct pfsync_header);
1200 case PFSYNC_ACT_UREQ:
1201 len = (sc->sc_maxcount * sizeof(struct pfsync_state_upd_req)) +
1202 sizeof(struct pfsync_header);
1204 case PFSYNC_ACT_BUS:
1205 len = sizeof(struct pfsync_header) +
1206 sizeof(struct pfsync_state_bus);
1208 case PFSYNC_ACT_TDB_UPD:
1209 len = (sc->sc_maxcount * sizeof(struct pfsync_tdb)) +
1210 sizeof(struct pfsync_header);
1213 len = (sc->sc_maxcount * sizeof(struct pfsync_state)) +
1214 sizeof(struct pfsync_header);
1219 MCLGET(m, M_WAITOK);
1220 if ((m->m_flags & M_EXT) == 0) {
1222 sc->sc_if.if_oerrors++;
1225 m->m_data += (MCLBYTES - len) &~ (sizeof(long) - 1);
1229 m->m_pkthdr.rcvif = NULL;
1230 m->m_pkthdr.len = m->m_len = sizeof(struct pfsync_header);
1231 h = mtod(m, struct pfsync_header *);
1232 h->version = PFSYNC_VERSION;
1237 *sp = (void *)((char *)h + PFSYNC_HDRLEN);
1238 lwkt_reltoken(&pf_token);
1239 callout_reset(&sc->sc_tmo, hz, pfsync_timeout,
1240 LIST_FIRST(&pfsync_list));
1241 lwkt_gettoken(&pf_token);
1246 pfsync_pack_state(u_int8_t action, struct pf_state *st, int flags)
1248 struct ifnet *ifp = NULL;
1249 struct pfsync_softc *sc = pfsyncif;
1250 struct pfsync_header *h, *h_net;
1251 struct pfsync_state *sp = NULL;
1252 struct pfsync_state_upd *up = NULL;
1253 struct pfsync_state_del *dp = NULL;
1255 u_int8_t i = 255, newaction = 0;
1262 * If a packet falls in the forest and there's nobody around to
1263 * hear, does it make a sound?
1265 if (ifp->if_bpf == NULL && sc->sc_sync_ifp == NULL &&
1266 sc->sc_sync_peer.s_addr == INADDR_PFSYNC_GROUP) {
1267 /* Don't leave any stale pfsync packets hanging around. */
1268 if (sc->sc_mbuf != NULL) {
1269 m_freem(sc->sc_mbuf);
1271 sc->sc_statep.s = NULL;
1276 if (action >= PFSYNC_ACT_MAX)
1280 if (sc->sc_mbuf == NULL) {
1281 if ((sc->sc_mbuf = pfsync_get_mbuf(sc, action,
1282 (void *)&sc->sc_statep.s)) == NULL) {
1286 h = mtod(sc->sc_mbuf, struct pfsync_header *);
1288 h = mtod(sc->sc_mbuf, struct pfsync_header *);
1289 if (h->action != action) {
1291 if ((sc->sc_mbuf = pfsync_get_mbuf(sc, action,
1292 (void *)&sc->sc_statep.s)) == NULL) {
1296 h = mtod(sc->sc_mbuf, struct pfsync_header *);
1299 * If it's an update, look in the packet to see if
1300 * we already have an update for the state.
1302 if (action == PFSYNC_ACT_UPD && sc->sc_maxupdates) {
1303 struct pfsync_state *usp =
1304 (void *)((char *)h + PFSYNC_HDRLEN);
1306 for (i = 0; i < h->count; i++) {
1307 if (!memcmp(usp->id, &st->id,
1309 usp->creatorid == st->creatorid) {
1320 st->pfsync_time = mycpu->gd_time_seconds;
1323 /* not a "duplicate" update */
1325 sp = sc->sc_statep.s++;
1326 sc->sc_mbuf->m_pkthdr.len =
1327 sc->sc_mbuf->m_len += sizeof(struct pfsync_state);
1329 bzero(sp, sizeof(*sp));
1331 pfsync_state_export(sp, st);
1333 if (flags & PFSYNC_FLAG_STALE)
1334 sp->sync_flags |= PFSTATE_STALE;
1336 pf_state_peer_hton(&st->src, &sp->src);
1337 pf_state_peer_hton(&st->dst, &sp->dst);
1339 if (st->expire <= time_second)
1340 sp->expire = htonl(0);
1342 sp->expire = htonl(st->expire - time_second);
1345 /* do we need to build "compressed" actions for network transfer? */
1346 if (sc->sc_sync_ifp && flags & PFSYNC_FLAG_COMPRESS) {
1348 case PFSYNC_ACT_UPD:
1349 newaction = PFSYNC_ACT_UPD_C;
1351 case PFSYNC_ACT_DEL:
1352 newaction = PFSYNC_ACT_DEL_C;
1355 /* by default we just send the uncompressed states */
1361 if (sc->sc_mbuf_net == NULL) {
1362 if ((sc->sc_mbuf_net = pfsync_get_mbuf(sc, newaction,
1363 (void *)&sc->sc_statep_net.s)) == NULL) {
1368 h_net = mtod(sc->sc_mbuf_net, struct pfsync_header *);
1370 switch (newaction) {
1371 case PFSYNC_ACT_UPD_C:
1373 up = (void *)((char *)h_net +
1374 PFSYNC_HDRLEN + (i * sizeof(*up)));
1378 sc->sc_mbuf_net->m_pkthdr.len =
1379 sc->sc_mbuf_net->m_len += sizeof(*up);
1380 up = sc->sc_statep_net.u++;
1382 bzero(up, sizeof(*up));
1383 bcopy(&st->id, up->id, sizeof(up->id));
1384 up->creatorid = st->creatorid;
1386 up->timeout = st->timeout;
1387 up->expire = sp->expire;
1391 case PFSYNC_ACT_DEL_C:
1392 sc->sc_mbuf_net->m_pkthdr.len =
1393 sc->sc_mbuf_net->m_len += sizeof(*dp);
1394 dp = sc->sc_statep_net.d++;
1397 bzero(dp, sizeof(*dp));
1398 bcopy(&st->id, dp->id, sizeof(dp->id));
1399 dp->creatorid = st->creatorid;
1404 if (h->count == sc->sc_maxcount ||
1405 (sc->sc_maxupdates && (sp->updates >= sc->sc_maxupdates)))
1406 ret = pfsync_sendout(sc);
1413 pfsync_request_update(struct pfsync_state_upd *up, struct in_addr *src)
1415 struct pfsync_header *h;
1416 struct pfsync_softc *sc = pfsyncif;
1417 struct pfsync_state_upd_req *rup;
1423 if (sc->sc_mbuf == NULL) {
1424 if ((sc->sc_mbuf = pfsync_get_mbuf(sc, PFSYNC_ACT_UREQ,
1425 (void *)&sc->sc_statep.s)) == NULL)
1427 h = mtod(sc->sc_mbuf, struct pfsync_header *);
1429 h = mtod(sc->sc_mbuf, struct pfsync_header *);
1430 if (h->action != PFSYNC_ACT_UREQ) {
1432 if ((sc->sc_mbuf = pfsync_get_mbuf(sc, PFSYNC_ACT_UREQ,
1433 (void *)&sc->sc_statep.s)) == NULL)
1435 h = mtod(sc->sc_mbuf, struct pfsync_header *);
1440 sc->sc_sendaddr = *src;
1441 sc->sc_mbuf->m_pkthdr.len = sc->sc_mbuf->m_len += sizeof(*rup);
1443 rup = sc->sc_statep.r++;
1444 bzero(rup, sizeof(*rup));
1446 bcopy(up->id, rup->id, sizeof(rup->id));
1447 rup->creatorid = up->creatorid;
1450 if (h->count == sc->sc_maxcount)
1451 ret = pfsync_sendout(sc);
1457 pfsync_clear_states(u_int32_t creatorid, char *ifname)
1459 struct pfsync_softc *sc = pfsyncif;
1460 struct pfsync_state_clr *cp;
1467 if (sc->sc_mbuf != NULL)
1469 if ((sc->sc_mbuf = pfsync_get_mbuf(sc, PFSYNC_ACT_CLR,
1470 (void *)&sc->sc_statep.c)) == NULL) {
1474 sc->sc_mbuf->m_pkthdr.len = sc->sc_mbuf->m_len += sizeof(*cp);
1475 cp = sc->sc_statep.c;
1476 cp->creatorid = creatorid;
1478 strlcpy(cp->ifname, ifname, IFNAMSIZ);
1480 ret = (pfsync_sendout(sc));
1486 pfsync_timeout(void *v)
1488 struct pfsync_softc *sc = v;
1496 pfsync_send_bus(struct pfsync_softc *sc, u_int8_t status)
1498 struct pfsync_state_bus *bus;
1500 if (sc->sc_mbuf != NULL)
1503 if (pfsync_sync_ok &&
1504 (sc->sc_mbuf = pfsync_get_mbuf(sc, PFSYNC_ACT_BUS,
1505 (void *)&sc->sc_statep.b)) != NULL) {
1506 sc->sc_mbuf->m_pkthdr.len = sc->sc_mbuf->m_len += sizeof(*bus);
1507 bus = sc->sc_statep.b;
1508 bus->creatorid = pf_status.hostid;
1509 bus->status = status;
1510 bus->endtime = htonl(mycpu->gd_time_seconds - sc->sc_ureq_received);
1516 pfsync_bulk_update(void *v)
1518 struct pfsync_softc *sc = v;
1520 struct pf_state *state;
1522 ASSERT_LWKT_TOKEN_HELD(&pf_token);
1525 if (sc->sc_mbuf != NULL)
1529 * Grab at most PFSYNC_BULKPACKETS worth of states which have not
1530 * been sent since the latest request was made.
1532 state = sc->sc_bulk_send_next;
1535 /* send state update if syncable and not already sent */
1536 if (!state->sync_flags
1537 && state->timeout < PFTM_MAX
1538 && state->pfsync_time <= sc->sc_ureq_received) {
1539 pfsync_pack_state(PFSYNC_ACT_UPD, state, 0);
1543 /* figure next state to send */
1544 state = TAILQ_NEXT(state, entry_list);
1546 /* wrap to start of list if we hit the end */
1548 state = TAILQ_FIRST(&state_list);
1549 } while (i < sc->sc_maxcount * PFSYNC_BULKPACKETS &&
1550 state != sc->sc_bulk_terminator);
1552 if (!state || state == sc->sc_bulk_terminator) {
1554 pfsync_send_bus(sc, PFSYNC_BUS_END);
1555 sc->sc_ureq_received = 0;
1556 sc->sc_bulk_send_next = NULL;
1557 sc->sc_bulk_terminator = NULL;
1558 lwkt_reltoken(&pf_token);
1559 callout_stop(&sc->sc_bulk_tmo);
1560 lwkt_gettoken(&pf_token);
1561 if (pf_status.debug >= PF_DEBUG_MISC)
1562 kprintf("pfsync: bulk update complete\n");
1564 /* look again for more in a bit */
1565 lwkt_reltoken(&pf_token);
1566 callout_reset(&sc->sc_bulk_tmo, 1, pfsync_timeout,
1567 LIST_FIRST(&pfsync_list));
1568 lwkt_gettoken(&pf_token);
1569 sc->sc_bulk_send_next = state;
1571 if (sc->sc_mbuf != NULL)
1577 pfsync_bulkfail(void *v)
1579 struct pfsync_softc *sc = v;
1582 ASSERT_LWKT_TOKEN_HELD(&pf_token);
1584 if (sc->sc_bulk_tries++ < PFSYNC_MAX_BULKTRIES) {
1585 /* Try again in a bit */
1586 lwkt_reltoken(&pf_token);
1587 callout_reset(&sc->sc_bulkfail_tmo, 5 * hz, pfsync_bulkfail,
1588 LIST_FIRST(&pfsync_list));
1589 lwkt_gettoken(&pf_token);
1591 error = pfsync_request_update(NULL, NULL);
1592 if (error == ENOMEM) {
1593 if (pf_status.debug >= PF_DEBUG_MISC)
1594 kprintf("pfsync: cannot allocate mbufs for "
1600 /* Pretend like the transfer was ok */
1601 sc->sc_ureq_sent = 0;
1602 sc->sc_bulk_tries = 0;
1604 if (!pfsync_sync_ok)
1605 carp_group_demote_adj(&sc->sc_if, -1);
1608 if (pf_status.debug >= PF_DEBUG_MISC)
1609 kprintf("pfsync: failed to receive "
1610 "bulk update status\n");
1611 lwkt_reltoken(&pf_token);
1612 callout_stop(&sc->sc_bulkfail_tmo);
1613 lwkt_gettoken(&pf_token);
1617 /* This must be called in splnet() */
1619 pfsync_sendout(struct pfsync_softc *sc)
1622 struct ifnet *ifp = &sc->sc_if;
1626 ASSERT_LWKT_TOKEN_HELD(&pf_token);
1628 lwkt_reltoken(&pf_token);
1629 callout_stop(&sc->sc_tmo);
1630 lwkt_gettoken(&pf_token);
1632 if (sc->sc_mbuf == NULL)
1636 sc->sc_statep.s = NULL;
1642 bpf_mtap(ifp->if_bpf, m, BPF_DIRECTION_OUT);
1647 if (sc->sc_mbuf_net) {
1649 m = sc->sc_mbuf_net;
1650 sc->sc_mbuf_net = NULL;
1651 sc->sc_statep_net.s = NULL;
1654 return pfsync_sendout_mbuf(sc, m);
1658 pfsync_sendout_mbuf(struct pfsync_softc *sc, struct mbuf *m)
1663 if (sc->sc_sync_ifp ||
1664 sc->sc_sync_peer.s_addr != INADDR_PFSYNC_GROUP) {
1665 M_PREPEND(m, sizeof(struct ip), M_WAITOK);
1667 pfsyncstats.pfsyncs_onomem++;
1670 ip = mtod(m, struct ip *);
1671 ip->ip_v = IPVERSION;
1672 ip->ip_hl = sizeof(*ip) >> 2;
1673 ip->ip_tos = IPTOS_LOWDELAY;
1674 ip->ip_len = htons(m->m_pkthdr.len);
1675 ip->ip_id = htons(ip_randomid());
1676 ip->ip_off = htons(IP_DF);
1677 ip->ip_ttl = PFSYNC_DFLTTL;
1678 ip->ip_p = IPPROTO_PFSYNC;
1681 bzero(&sa, sizeof(sa));
1682 ip->ip_src.s_addr = INADDR_ANY;
1684 if (sc->sc_sendaddr.s_addr == INADDR_PFSYNC_GROUP)
1685 m->m_flags |= M_MCAST;
1686 ip->ip_dst = sc->sc_sendaddr;
1687 sc->sc_sendaddr.s_addr = sc->sc_sync_peer.s_addr;
1689 pfsyncstats.pfsyncs_opackets++;
1691 if (ip_output(m, NULL, NULL, IP_RAWOUTPUT, &sc->sc_imo, NULL))
1692 pfsyncstats.pfsyncs_oerrors++;
1700 pfsync_modevent(module_t mod, int type, void *data)
1704 struct pfsync_softc *pfs_if, *tmp;
1706 lwkt_gettoken(&pf_token);
1710 LIST_INIT(&pfsync_list);
1711 lwkt_reltoken(&pf_token);
1712 if_clone_attach(&pfsync_cloner);
1713 lwkt_gettoken(&pf_token);
1714 /* Override the function pointer for pf_ioctl.c */
1718 lwkt_reltoken(&pf_token);
1719 if_clone_detach(&pfsync_cloner);
1720 lwkt_gettoken(&pf_token);
1721 LIST_FOREACH_MUTABLE(pfs_if, &pfsync_list, sc_next, tmp) {
1722 pfsync_clone_destroy(&pfs_if->sc_if);
1731 lwkt_reltoken(&pf_token);
1735 static moduledata_t pfsync_mod = {
1741 #define PFSYNC_MODVER 44
1743 DECLARE_MODULE(pfsync, pfsync_mod, SI_SUB_PSEUDO, SI_ORDER_ANY);
1744 MODULE_VERSION(pfsync, PFSYNC_MODVER);