1 /* $NetBSD: if_pfsync.c,v 1.23 2024/07/05 04:31:52 rin Exp $ */
2 /* $OpenBSD: if_pfsync.c,v 1.83 2007/06/26 14:44:12 mcbride Exp $ */
3
4 /*
5 * Copyright (c) 2002 Michael Shalayeff
6 * All rights reserved.
7 *
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
10 * are met:
11 * 1. Redistributions of source code must retain the above copyright
12 * notice, this list of conditions and the following disclaimer.
13 * 2. Redistributions in binary form must reproduce the above copyright
14 * notice, this list of conditions and the following disclaimer in the
15 * documentation and/or other materials provided with the distribution.
16 *
17 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
18 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
19 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
20 * IN NO EVENT SHALL THE AUTHOR OR HIS RELATIVES BE LIABLE FOR ANY DIRECT,
21 * INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
22 * (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
23 * SERVICES; LOSS OF MIND, USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT,
25 * STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING
26 * IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF
27 * THE POSSIBILITY OF SUCH DAMAGE.
28 */
29
30 #include <sys/cdefs.h>
31 __KERNEL_RCSID(0, "$NetBSD: if_pfsync.c,v 1.23 2024/07/05 04:31:52 rin Exp $");
32
33 #ifdef _KERNEL_OPT
34 #include "opt_inet.h"
35 #include "opt_inet6.h"
36 #endif
37
38 #include <sys/param.h>
39 #include <sys/proc.h>
40 #include <sys/systm.h>
41 #include <sys/time.h>
42 #include <sys/mbuf.h>
43 #include <sys/socket.h>
44 #include <sys/ioctl.h>
45 #include <sys/callout.h>
46 #include <sys/kernel.h>
47
48 #include <net/if.h>
49 #include <net/if_types.h>
50 #include <net/route.h>
51 #include <net/bpf.h>
52 #include <netinet/in.h>
53 #ifndef __NetBSD__
54 #include <netinet/if_ether.h>
55 #else
56 #include <net/if_ether.h>
57 #endif /* __NetBSD__ */
58 #include <netinet/tcp.h>
59 #include <netinet/tcp_seq.h>
60
61 #ifdef INET
62 #include <netinet/in_systm.h>
63 #include <netinet/in_var.h>
64 #include <netinet/ip.h>
65 #include <netinet/ip_var.h>
66 #endif
67
68 #ifdef INET6
69 #include <netinet6/nd6.h>
70 #endif /* INET6 */
71
72 #include "carp.h"
73 #if NCARP > 0
74 extern int carp_suppress_preempt;
75 #endif
76
77 #include <net/pfvar.h>
78 #include <net/if_pfsync.h>
79
80 #ifdef __NetBSD__
81 #include <sys/conf.h>
82 #include <sys/lwp.h>
83 #include <sys/kauth.h>
84 #include <sys/sysctl.h>
85
86 #include <net/net_stats.h>
87
88 #include "ioconf.h"
89
90 percpu_t *pfsyncstat_percpu;
91
92 #define PFSYNC_STATINC(x) _NET_STATINC(pfsyncstat_percpu, x)
93 #endif /* __NetBSD__ */
94
95 #include "pfsync.h"
96
97 #define PFSYNC_MINMTU \
98 (sizeof(struct pfsync_header) + sizeof(struct pf_state))
99
100 #ifdef PFSYNCDEBUG
101 #define DPRINTF(x) do { if (pfsyncdebug) printf x ; } while (0)
102 int pfsyncdebug;
103 #else
104 #define DPRINTF(x)
105 #endif
106
107 extern int ifqmaxlen; /* XXX */
108
109 struct pfsync_softc *pfsyncif = NULL;
110
111 int pfsync_clone_create(struct if_clone *, int);
112 int pfsync_clone_destroy(struct ifnet *);
113 void pfsync_setmtu(struct pfsync_softc *, int);
114 int pfsync_alloc_scrub_memory(struct pfsync_state_peer *,
115 struct pf_state_peer *);
116 int pfsync_insert_net_state(struct pfsync_state *, u_int8_t);
117 void pfsync_update_net_tdb(struct pfsync_tdb *);
118 int pfsyncoutput(struct ifnet *, struct mbuf *, const struct sockaddr *,
119 const struct rtentry *);
120 int pfsyncioctl(struct ifnet *, u_long, void*);
121 void pfsyncstart(struct ifnet *);
122
123 struct mbuf *pfsync_get_mbuf(struct pfsync_softc *, u_int8_t, void **);
124 int pfsync_request_update(struct pfsync_state_upd *, struct in_addr *);
125 int pfsync_sendout(struct pfsync_softc *);
126 int pfsync_tdb_sendout(struct pfsync_softc *);
127 int pfsync_sendout_mbuf(struct pfsync_softc *, struct mbuf *);
128 void pfsync_timeout(void *);
129 void pfsync_tdb_timeout(void *);
130 void pfsync_send_bus(struct pfsync_softc *, u_int8_t);
131 void pfsync_bulk_update(void *);
132 void pfsync_bulkfail(void *);
133
134 int pfsync_sync_ok;
135
136 struct if_clone pfsync_cloner =
137 IF_CLONE_INITIALIZER("pfsync", pfsync_clone_create, pfsync_clone_destroy);
138
139 void
pfsyncattach(int npfsync)140 pfsyncattach(int npfsync)
141 {
142 if_clone_attach(&pfsync_cloner);
143
144 pfsyncstat_percpu = percpu_alloc(sizeof(uint64_t) * PFSYNC_NSTATS);
145 }
146
147 int
pfsync_clone_create(struct if_clone * ifc,int unit)148 pfsync_clone_create(struct if_clone *ifc, int unit)
149 {
150 struct ifnet *ifp;
151
152 if (unit != 0)
153 return (EINVAL);
154
155 pfsync_sync_ok = 1;
156 if ((pfsyncif = malloc(sizeof(*pfsyncif), M_DEVBUF, M_NOWAIT)) == NULL)
157 return (ENOMEM);
158 memset(pfsyncif, 0, sizeof(*pfsyncif));
159 pfsyncif->sc_mbuf = NULL;
160 pfsyncif->sc_mbuf_net = NULL;
161 pfsyncif->sc_mbuf_tdb = NULL;
162 pfsyncif->sc_statep.s = NULL;
163 pfsyncif->sc_statep_net.s = NULL;
164 pfsyncif->sc_statep_tdb.t = NULL;
165 pfsyncif->sc_maxupdates = 128;
166 pfsyncif->sc_sync_peer.s_addr = INADDR_PFSYNC_GROUP;
167 pfsyncif->sc_sendaddr.s_addr = INADDR_PFSYNC_GROUP;
168 pfsyncif->sc_ureq_received = 0;
169 pfsyncif->sc_ureq_sent = 0;
170 pfsyncif->sc_bulk_send_next = NULL;
171 pfsyncif->sc_bulk_terminator = NULL;
172 ifp = &pfsyncif->sc_if;
173 snprintf(ifp->if_xname, sizeof ifp->if_xname, "pfsync%d", unit);
174 ifp->if_softc = pfsyncif;
175 ifp->if_ioctl = pfsyncioctl;
176 ifp->if_output = pfsyncoutput;
177 ifp->if_start = pfsyncstart;
178 ifp->if_type = IFT_PFSYNC;
179 ifp->if_snd.ifq_maxlen = ifqmaxlen;
180 ifp->if_hdrlen = PFSYNC_HDRLEN;
181 pfsync_setmtu(pfsyncif, ETHERMTU);
182
183 callout_init(&pfsyncif->sc_tmo, 0);
184 callout_init(&pfsyncif->sc_tdb_tmo, 0);
185 callout_init(&pfsyncif->sc_bulk_tmo, 0);
186 callout_init(&pfsyncif->sc_bulkfail_tmo, 0);
187 callout_setfunc(&pfsyncif->sc_tmo, pfsync_timeout, pfsyncif);
188 callout_setfunc(&pfsyncif->sc_tdb_tmo, pfsync_tdb_timeout, pfsyncif);
189 callout_setfunc(&pfsyncif->sc_bulk_tmo, pfsync_bulk_update, pfsyncif);
190 callout_setfunc(&pfsyncif->sc_bulkfail_tmo, pfsync_bulkfail, pfsyncif);
191
192 if_attach(ifp);
193 if_alloc_sadl(ifp);
194
195 bpf_attach(&pfsyncif->sc_if, DLT_PFSYNC, PFSYNC_HDRLEN);
196
197 return (0);
198 }
199
200 int
pfsync_clone_destroy(struct ifnet * ifp)201 pfsync_clone_destroy(struct ifnet *ifp)
202 {
203 bpf_detach(ifp);
204 if_detach(ifp);
205 free(pfsyncif, M_DEVBUF);
206 pfsyncif = NULL;
207 return (0);
208 }
209
210 /*
211 * Start output on the pfsync interface.
212 */
213 void
pfsyncstart(struct ifnet * ifp)214 pfsyncstart(struct ifnet *ifp)
215 {
216 struct mbuf *m;
217 int s;
218
219 for (;;) {
220 s = splnet();
221 IF_DROP(&ifp->if_snd);
222 IF_DEQUEUE(&ifp->if_snd, m);
223 splx(s);
224
225 if (m == NULL)
226 return;
227 else
228 m_freem(m);
229 }
230 }
231
232 int
pfsync_alloc_scrub_memory(struct pfsync_state_peer * s,struct pf_state_peer * d)233 pfsync_alloc_scrub_memory(struct pfsync_state_peer *s,
234 struct pf_state_peer *d)
235 {
236 if (s->scrub.scrub_flag && d->scrub == NULL) {
237 d->scrub = pool_get(&pf_state_scrub_pl, PR_NOWAIT);
238 if (d->scrub == NULL)
239 return (ENOMEM);
240 memset(d->scrub, 0, sizeof(*d->scrub));
241 }
242
243 return (0);
244 }
245
246 int
pfsync_insert_net_state(struct pfsync_state * sp,u_int8_t chksum_flag)247 pfsync_insert_net_state(struct pfsync_state *sp, u_int8_t chksum_flag)
248 {
249 struct pf_state *st = NULL;
250 struct pf_state_key *sk = NULL;
251 struct pf_rule *r = NULL;
252 struct pfi_kif *kif;
253
254 if (sp->creatorid == 0 && pf_status.debug >= PF_DEBUG_MISC) {
255 printf("pfsync_insert_net_state: invalid creator id:"
256 " %08x\n", ntohl(sp->creatorid));
257 return (EINVAL);
258 }
259
260 kif = pfi_kif_get(sp->ifname);
261 if (kif == NULL) {
262 if (pf_status.debug >= PF_DEBUG_MISC)
263 printf("pfsync_insert_net_state: "
264 "unknown interface: %s\n", sp->ifname);
265 /* skip this state */
266 return (0);
267 }
268
269 /*
270 * If the ruleset checksums match, it's safe to associate the state
271 * with the rule of that number.
272 */
273 if (sp->rule != htonl(-1) && sp->anchor == htonl(-1) && chksum_flag &&
274 ntohl(sp->rule) <
275 pf_main_ruleset.rules[PF_RULESET_FILTER].active.rcount)
276 r = pf_main_ruleset.rules[
277 PF_RULESET_FILTER].active.ptr_array[ntohl(sp->rule)];
278 else
279 r = &pf_default_rule;
280
281 if (!r->max_states || r->states < r->max_states)
282 st = pool_get(&pf_state_pl, PR_NOWAIT);
283 if (st == NULL) {
284 pfi_kif_unref(kif, PFI_KIF_REF_NONE);
285 return (ENOMEM);
286 }
287 memset(st, 0, sizeof(*st));
288
289 if ((sk = pf_alloc_state_key(st)) == NULL) {
290 pool_put(&pf_state_pl, st);
291 pfi_kif_unref(kif, PFI_KIF_REF_NONE);
292 return (ENOMEM);
293 }
294
295 /* allocate memory for scrub info */
296 if (pfsync_alloc_scrub_memory(&sp->src, &st->src) ||
297 pfsync_alloc_scrub_memory(&sp->dst, &st->dst)) {
298 pfi_kif_unref(kif, PFI_KIF_REF_NONE);
299 if (st->src.scrub)
300 pool_put(&pf_state_scrub_pl, st->src.scrub);
301 pool_put(&pf_state_pl, st);
302 pool_put(&pf_state_key_pl, sk);
303 return (ENOMEM);
304 }
305
306 st->rule.ptr = r;
307 /* XXX get pointers to nat_rule and anchor */
308
309 /* XXX when we have nat_rule/anchors, use STATE_INC_COUNTERS */
310 r->states++;
311
312 /* fill in the rest of the state entry */
313 pf_state_host_ntoh(&sp->lan, &sk->lan);
314 pf_state_host_ntoh(&sp->gwy, &sk->gwy);
315 pf_state_host_ntoh(&sp->ext, &sk->ext);
316
317 pf_state_peer_ntoh(&sp->src, &st->src);
318 pf_state_peer_ntoh(&sp->dst, &st->dst);
319
320 memcpy(&st->rt_addr, &sp->rt_addr, sizeof(st->rt_addr));
321 st->creation = time_second - ntohl(sp->creation);
322 st->expire = ntohl(sp->expire) + time_second;
323
324 sk->af = sp->af;
325 sk->proto = sp->proto;
326 sk->direction = sp->direction;
327 st->log = sp->log;
328 st->timeout = sp->timeout;
329 st->allow_opts = sp->allow_opts;
330
331 memcpy(&st->id, sp->id, sizeof(st->id));
332 st->creatorid = sp->creatorid;
333 st->sync_flags = PFSTATE_FROMSYNC;
334
335 if (pf_insert_state(kif, st)) {
336 pfi_kif_unref(kif, PFI_KIF_REF_NONE);
337 /* XXX when we have nat_rule/anchors, use STATE_DEC_COUNTERS */
338 r->states--;
339 if (st->dst.scrub)
340 pool_put(&pf_state_scrub_pl, st->dst.scrub);
341 if (st->src.scrub)
342 pool_put(&pf_state_scrub_pl, st->src.scrub);
343 pool_put(&pf_state_pl, st);
344 return (EINVAL);
345 }
346
347 return (0);
348 }
349
350 void
pfsync_input(struct mbuf * m,int off,int proto)351 pfsync_input(struct mbuf *m, int off, int proto)
352 {
353 struct ip *ip = mtod(m, struct ip *);
354 struct pfsync_header *ph;
355 struct pfsync_softc *sc = pfsyncif;
356 struct pf_state *st;
357 struct pf_state_key *sk;
358 struct pf_state_cmp id_key;
359 struct pfsync_state *sp;
360 struct pfsync_state_upd *up;
361 struct pfsync_state_del *dp;
362 struct pfsync_state_clr *cp;
363 struct pfsync_state_upd_req *rup;
364 struct pfsync_state_bus *bus;
365 struct in_addr src;
366 struct mbuf *mp;
367 int iplen, action, error, i, s, count, offp, sfail, stale = 0;
368 u_int8_t chksum_flag = 0;
369
370 PFSYNC_STATINC(PFSYNC_STAT_IPACKETS);
371
372 /* verify that we have a sync interface configured */
373 if (!sc || !sc->sc_sync_ifp || !pf_status.running)
374 goto done;
375
376 /* verify that the packet came in on the right interface */
377 if (sc->sc_sync_ifp->if_index != m->m_pkthdr.rcvif_index) {
378 PFSYNC_STATINC(PFSYNC_STAT_BADIF);
379 goto done;
380 }
381
382 /* verify that the IP TTL is 255. */
383 if (ip->ip_ttl != PFSYNC_DFLTTL) {
384 PFSYNC_STATINC(PFSYNC_STAT_BADTTL);
385 goto done;
386 }
387
388 iplen = ip->ip_hl << 2;
389
390 if (m->m_pkthdr.len < iplen + sizeof(*ph)) {
391 PFSYNC_STATINC(PFSYNC_STAT_HDROPS);
392 goto done;
393 }
394
395 if (iplen + sizeof(*ph) > m->m_len) {
396 if ((m = m_pullup(m, iplen + sizeof(*ph))) == NULL) {
397 PFSYNC_STATINC(PFSYNC_STAT_HDROPS);
398 goto done;
399 }
400 ip = mtod(m, struct ip *);
401 }
402 ph = (struct pfsync_header *)((char *)ip + iplen);
403
404 /* verify the version */
405 if (ph->version != PFSYNC_VERSION) {
406 PFSYNC_STATINC(PFSYNC_STAT_BADVER);
407 goto done;
408 }
409
410 action = ph->action;
411 count = ph->count;
412
413 /* make sure it's a valid action code */
414 if (action >= PFSYNC_ACT_MAX) {
415 PFSYNC_STATINC(PFSYNC_STAT_BADACT);
416 goto done;
417 }
418
419 /* Cheaper to grab this now than having to mess with mbufs later */
420 src = ip->ip_src;
421
422 if (!bcmp(&ph->pf_chksum, &pf_status.pf_chksum, PF_MD5_DIGEST_LENGTH))
423 chksum_flag++;
424
425 switch (action) {
426 case PFSYNC_ACT_CLR: {
427 struct pf_state *nexts;
428 struct pf_state_key *nextsk;
429 struct pfi_kif *kif;
430 u_int32_t creatorid;
431 if ((mp = m_pulldown(m, iplen + sizeof(*ph),
432 sizeof(*cp), &offp)) == NULL) {
433 PFSYNC_STATINC(PFSYNC_STAT_BADLEN);
434 return;
435 }
436 cp = (struct pfsync_state_clr *)(mp->m_data + offp);
437 creatorid = cp->creatorid;
438
439 s = splsoftnet();
440 if (cp->ifname[0] == '\0') {
441 for (st = RB_MIN(pf_state_tree_id, &tree_id);
442 st; st = nexts) {
443 nexts = RB_NEXT(pf_state_tree_id, &tree_id, st);
444 if (st->creatorid == creatorid) {
445 st->sync_flags |= PFSTATE_FROMSYNC;
446 pf_unlink_state(st);
447 }
448 }
449 } else {
450 if ((kif = pfi_kif_get(cp->ifname)) == NULL) {
451 splx(s);
452 return;
453 }
454 for (sk = RB_MIN(pf_state_tree_lan_ext,
455 &pf_statetbl_lan_ext); sk; sk = nextsk) {
456 nextsk = RB_NEXT(pf_state_tree_lan_ext,
457 &pf_statetbl_lan_ext, sk);
458 TAILQ_FOREACH(st, &sk->states, next) {
459 if (st->creatorid == creatorid) {
460 st->sync_flags |=
461 PFSTATE_FROMSYNC;
462 pf_unlink_state(st);
463 }
464 }
465 }
466 }
467 splx(s);
468
469 break;
470 }
471 case PFSYNC_ACT_INS:
472 if ((mp = m_pulldown(m, iplen + sizeof(*ph),
473 count * sizeof(*sp), &offp)) == NULL) {
474 PFSYNC_STATINC(PFSYNC_STAT_BADLEN);
475 return;
476 }
477
478 s = splsoftnet();
479 for (i = 0, sp = (struct pfsync_state *)(mp->m_data + offp);
480 i < count; i++, sp++) {
481 /* check for invalid values */
482 if (sp->timeout >= PFTM_MAX ||
483 sp->src.state > PF_TCPS_PROXY_DST ||
484 sp->dst.state > PF_TCPS_PROXY_DST ||
485 sp->direction > PF_OUT ||
486 (sp->af != AF_INET && sp->af != AF_INET6)) {
487 if (pf_status.debug >= PF_DEBUG_MISC)
488 printf("pfsync_insert: PFSYNC_ACT_INS: "
489 "invalid value\n");
490 PFSYNC_STATINC(PFSYNC_STAT_BADSTATE);
491 continue;
492 }
493
494 if ((error = pfsync_insert_net_state(sp,
495 chksum_flag))) {
496 if (error == ENOMEM) {
497 splx(s);
498 goto done;
499 }
500 continue;
501 }
502 }
503 splx(s);
504 break;
505 case PFSYNC_ACT_UPD:
506 if ((mp = m_pulldown(m, iplen + sizeof(*ph),
507 count * sizeof(*sp), &offp)) == NULL) {
508 PFSYNC_STATINC(PFSYNC_STAT_BADLEN);
509 return;
510 }
511
512 s = splsoftnet();
513 for (i = 0, sp = (struct pfsync_state *)(mp->m_data + offp);
514 i < count; i++, sp++) {
515 int flags = PFSYNC_FLAG_STALE;
516
517 /* check for invalid values */
518 if (sp->timeout >= PFTM_MAX ||
519 sp->src.state > PF_TCPS_PROXY_DST ||
520 sp->dst.state > PF_TCPS_PROXY_DST) {
521 if (pf_status.debug >= PF_DEBUG_MISC)
522 printf("pfsync_insert: PFSYNC_ACT_UPD: "
523 "invalid value\n");
524 PFSYNC_STATINC(PFSYNC_STAT_BADSTATE);
525 continue;
526 }
527
528 memcpy(&id_key.id, sp->id, sizeof(id_key.id));
529 id_key.creatorid = sp->creatorid;
530
531 st = pf_find_state_byid(&id_key);
532 if (st == NULL) {
533 /* insert the update */
534 if (pfsync_insert_net_state(sp, chksum_flag)) {
535 PFSYNC_STATINC(PFSYNC_STAT_BADSTATE);
536 }
537 continue;
538 }
539 sk = st->state_key;
540 sfail = 0;
541 if (sk->proto == IPPROTO_TCP) {
542 /*
543 * The state should never go backwards except
544 * for syn-proxy states. Neither should the
545 * sequence window slide backwards.
546 */
547 if (st->src.state > sp->src.state &&
548 (st->src.state < PF_TCPS_PROXY_SRC ||
549 sp->src.state >= PF_TCPS_PROXY_SRC))
550 sfail = 1;
551 else if (SEQ_GT(st->src.seqlo,
552 ntohl(sp->src.seqlo)))
553 sfail = 3;
554 else if (st->dst.state > sp->dst.state) {
555 /* There might still be useful
556 * information about the src state here,
557 * so import that part of the update,
558 * then "fail" so we send the updated
559 * state back to the peer who is missing
560 * our what we know. */
561 pf_state_peer_ntoh(&sp->src, &st->src);
562 /* XXX do anything with timeouts? */
563 sfail = 7;
564 flags = 0;
565 } else if (st->dst.state >= TCPS_SYN_SENT &&
566 SEQ_GT(st->dst.seqlo, ntohl(sp->dst.seqlo)))
567 sfail = 4;
568 } else {
569 /*
570 * Non-TCP protocol state machine always go
571 * forwards
572 */
573 if (st->src.state > sp->src.state)
574 sfail = 5;
575 else if (st->dst.state > sp->dst.state)
576 sfail = 6;
577 }
578 if (sfail) {
579 if (pf_status.debug >= PF_DEBUG_MISC)
580 printf("pfsync: %s stale update "
581 "(%d) id: %016" PRIu64 ""
582 "creatorid: %08x\n",
583 (sfail < 7 ? "ignoring"
584 : "partial"), sfail,
585 be64toh(st->id),
586 ntohl(st->creatorid));
587 PFSYNC_STATINC(PFSYNC_STAT_BADSTATE);
588
589 if (!(sp->sync_flags & PFSTATE_STALE)) {
590 /* we have a better state, send it */
591 if (sc->sc_mbuf != NULL && !stale)
592 pfsync_sendout(sc);
593 stale++;
594 if (!st->sync_flags)
595 pfsync_pack_state(
596 PFSYNC_ACT_UPD, st, flags);
597 }
598 continue;
599 }
600 pfsync_alloc_scrub_memory(&sp->dst, &st->dst);
601 pf_state_peer_ntoh(&sp->src, &st->src);
602 pf_state_peer_ntoh(&sp->dst, &st->dst);
603 st->expire = ntohl(sp->expire) + time_second;
604 st->timeout = sp->timeout;
605 }
606 if (stale && sc->sc_mbuf != NULL)
607 pfsync_sendout(sc);
608 splx(s);
609 break;
610 /*
611 * It's not strictly necessary for us to support the "uncompressed"
612 * delete action, but it's relatively simple and maintains consistency.
613 */
614 case PFSYNC_ACT_DEL:
615 if ((mp = m_pulldown(m, iplen + sizeof(*ph),
616 count * sizeof(*sp), &offp)) == NULL) {
617 PFSYNC_STATINC(PFSYNC_STAT_BADLEN);
618 return;
619 }
620
621 s = splsoftnet();
622 for (i = 0, sp = (struct pfsync_state *)(mp->m_data + offp);
623 i < count; i++, sp++) {
624 memcpy(&id_key.id, sp->id, sizeof(id_key.id));
625 id_key.creatorid = sp->creatorid;
626
627 st = pf_find_state_byid(&id_key);
628 if (st == NULL) {
629 PFSYNC_STATINC(PFSYNC_STAT_BADSTATE);
630 continue;
631 }
632 st->sync_flags |= PFSTATE_FROMSYNC;
633 pf_unlink_state(st);
634 }
635 splx(s);
636 break;
637 case PFSYNC_ACT_UPD_C: {
638 int update_requested = 0;
639
640 if ((mp = m_pulldown(m, iplen + sizeof(*ph),
641 count * sizeof(*up), &offp)) == NULL) {
642 PFSYNC_STATINC(PFSYNC_STAT_BADLEN);
643 return;
644 }
645
646 s = splsoftnet();
647 for (i = 0, up = (struct pfsync_state_upd *)(mp->m_data + offp);
648 i < count; i++, up++) {
649 /* check for invalid values */
650 if (up->timeout >= PFTM_MAX ||
651 up->src.state > PF_TCPS_PROXY_DST ||
652 up->dst.state > PF_TCPS_PROXY_DST) {
653 if (pf_status.debug >= PF_DEBUG_MISC)
654 printf("pfsync_insert: "
655 "PFSYNC_ACT_UPD_C: "
656 "invalid value\n");
657 PFSYNC_STATINC(PFSYNC_STAT_BADSTATE);
658 continue;
659 }
660
661 memcpy(&id_key.id, up->id, sizeof(id_key.id));
662 id_key.creatorid = up->creatorid;
663
664 st = pf_find_state_byid(&id_key);
665 if (st == NULL) {
666 /* We don't have this state. Ask for it. */
667 error = pfsync_request_update(up, &src);
668 if (error == ENOMEM) {
669 splx(s);
670 goto done;
671 }
672 update_requested = 1;
673 PFSYNC_STATINC(PFSYNC_STAT_BADSTATE);
674 continue;
675 }
676 sk = st->state_key;
677 sfail = 0;
678 if (sk->proto == IPPROTO_TCP) {
679 /*
680 * The state should never go backwards except
681 * for syn-proxy states. Neither should the
682 * sequence window slide backwards.
683 */
684 if (st->src.state > up->src.state &&
685 (st->src.state < PF_TCPS_PROXY_SRC ||
686 up->src.state >= PF_TCPS_PROXY_SRC))
687 sfail = 1;
688 else if (st->dst.state > up->dst.state)
689 sfail = 2;
690 else if (SEQ_GT(st->src.seqlo,
691 ntohl(up->src.seqlo)))
692 sfail = 3;
693 else if (st->dst.state >= TCPS_SYN_SENT &&
694 SEQ_GT(st->dst.seqlo, ntohl(up->dst.seqlo)))
695 sfail = 4;
696 } else {
697 /*
698 * Non-TCP protocol state machine always go
699 * forwards
700 */
701 if (st->src.state > up->src.state)
702 sfail = 5;
703 else if (st->dst.state > up->dst.state)
704 sfail = 6;
705 }
706 if (sfail) {
707 if (pf_status.debug >= PF_DEBUG_MISC)
708 printf("pfsync: ignoring stale update "
709 "(%d) id: %016" PRIu64 ""
710 "creatorid: %08x\n", sfail,
711 be64toh(st->id),
712 ntohl(st->creatorid));
713 PFSYNC_STATINC(PFSYNC_STAT_BADSTATE);
714
715 /* we have a better state, send it out */
716 if ((!stale || update_requested) &&
717 sc->sc_mbuf != NULL) {
718 pfsync_sendout(sc);
719 update_requested = 0;
720 }
721 stale++;
722 if (!st->sync_flags)
723 pfsync_pack_state(PFSYNC_ACT_UPD, st,
724 PFSYNC_FLAG_STALE);
725 continue;
726 }
727 pfsync_alloc_scrub_memory(&up->dst, &st->dst);
728 pf_state_peer_ntoh(&up->src, &st->src);
729 pf_state_peer_ntoh(&up->dst, &st->dst);
730 st->expire = ntohl(up->expire) + time_second;
731 st->timeout = up->timeout;
732 }
733 if ((update_requested || stale) && sc->sc_mbuf)
734 pfsync_sendout(sc);
735 splx(s);
736 break;
737 }
738 case PFSYNC_ACT_DEL_C:
739 if ((mp = m_pulldown(m, iplen + sizeof(*ph),
740 count * sizeof(*dp), &offp)) == NULL) {
741 PFSYNC_STATINC(PFSYNC_STAT_BADLEN);
742 return;
743 }
744
745 s = splsoftnet();
746 for (i = 0, dp = (struct pfsync_state_del *)(mp->m_data + offp);
747 i < count; i++, dp++) {
748 memcpy(&id_key.id, dp->id, sizeof(id_key.id));
749 id_key.creatorid = dp->creatorid;
750
751 st = pf_find_state_byid(&id_key);
752 if (st == NULL) {
753 PFSYNC_STATINC(PFSYNC_STAT_BADSTATE);
754 continue;
755 }
756 st->sync_flags |= PFSTATE_FROMSYNC;
757 pf_unlink_state(st);
758 }
759 splx(s);
760 break;
761 case PFSYNC_ACT_INS_F:
762 case PFSYNC_ACT_DEL_F:
763 /* not implemented */
764 break;
765 case PFSYNC_ACT_UREQ:
766 if ((mp = m_pulldown(m, iplen + sizeof(*ph),
767 count * sizeof(*rup), &offp)) == NULL) {
768 PFSYNC_STATINC(PFSYNC_STAT_BADLEN);
769 return;
770 }
771
772 s = splsoftnet();
773 if (sc->sc_mbuf != NULL)
774 pfsync_sendout(sc);
775 for (i = 0,
776 rup = (struct pfsync_state_upd_req *)(mp->m_data + offp);
777 i < count; i++, rup++) {
778 memcpy(&id_key.id, rup->id, sizeof(id_key.id));
779 id_key.creatorid = rup->creatorid;
780
781 if (id_key.id == 0 && id_key.creatorid == 0) {
782 sc->sc_ureq_received = time_uptime;
783 if (sc->sc_bulk_send_next == NULL)
784 sc->sc_bulk_send_next =
785 TAILQ_FIRST(&state_list);
786 sc->sc_bulk_terminator = sc->sc_bulk_send_next;
787 if (pf_status.debug >= PF_DEBUG_MISC)
788 printf("pfsync: received "
789 "bulk update request\n");
790 pfsync_send_bus(sc, PFSYNC_BUS_START);
791 callout_schedule(&sc->sc_bulk_tmo, 1 * hz);
792 } else {
793 st = pf_find_state_byid(&id_key);
794 if (st == NULL) {
795 PFSYNC_STATINC(PFSYNC_STAT_BADSTATE);
796 continue;
797 }
798 if (!st->sync_flags)
799 pfsync_pack_state(PFSYNC_ACT_UPD,
800 st, 0);
801 }
802 }
803 if (sc->sc_mbuf != NULL)
804 pfsync_sendout(sc);
805 splx(s);
806 break;
807 case PFSYNC_ACT_BUS:
808 /* If we're not waiting for a bulk update, who cares. */
809 if (sc->sc_ureq_sent == 0)
810 break;
811
812 if ((mp = m_pulldown(m, iplen + sizeof(*ph),
813 sizeof(*bus), &offp)) == NULL) {
814 PFSYNC_STATINC(PFSYNC_STAT_BADLEN);
815 return;
816 }
817 bus = (struct pfsync_state_bus *)(mp->m_data + offp);
818 switch (bus->status) {
819 case PFSYNC_BUS_START:
820 callout_schedule(&sc->sc_bulkfail_tmo,
821 pf_pool_limits[PF_LIMIT_STATES].limit /
822 (PFSYNC_BULKPACKETS * sc->sc_maxcount));
823 if (pf_status.debug >= PF_DEBUG_MISC)
824 printf("pfsync: received bulk "
825 "update start\n");
826 break;
827 case PFSYNC_BUS_END:
828 if (time_uptime - ntohl(bus->endtime) >=
829 sc->sc_ureq_sent) {
830 /* that's it, we're happy */
831 sc->sc_ureq_sent = 0;
832 sc->sc_bulk_tries = 0;
833 callout_stop(&sc->sc_bulkfail_tmo);
834 #if NCARP > 0
835 if (!pfsync_sync_ok)
836 carp_suppress_preempt--;
837 #endif
838 pfsync_sync_ok = 1;
839 if (pf_status.debug >= PF_DEBUG_MISC)
840 printf("pfsync: received valid "
841 "bulk update end\n");
842 } else {
843 if (pf_status.debug >= PF_DEBUG_MISC)
844 printf("pfsync: received invalid "
845 "bulk update end: bad timestamp\n");
846 }
847 break;
848 }
849 break;
850 }
851
852 done:
853 m_freem(m);
854 }
855
856 int
pfsyncoutput(struct ifnet * ifp,struct mbuf * m,const struct sockaddr * dst,const struct rtentry * rt)857 pfsyncoutput(struct ifnet *ifp, struct mbuf *m, const struct sockaddr *dst,
858 const struct rtentry *rt)
859 {
860 m_freem(m);
861 return (0);
862 }
863
864 /* ARGSUSED */
865 int
pfsyncioctl(struct ifnet * ifp,u_long cmd,void * data)866 pfsyncioctl(struct ifnet *ifp, u_long cmd, void* data)
867 {
868 struct lwp *l = curlwp;
869 struct pfsync_softc *sc = ifp->if_softc;
870 struct ifreq *ifr = (struct ifreq *)data;
871 struct ip_moptions *imo = &sc->sc_imo;
872 struct pfsyncreq pfsyncr;
873 struct ifnet *sifp;
874 int s, error;
875
876 switch (cmd) {
877 case SIOCSIFADDR:
878 case SIOCAIFADDR:
879 case SIOCSIFDSTADDR:
880 case SIOCSIFFLAGS:
881 if (ifp->if_flags & IFF_UP)
882 ifp->if_flags |= IFF_RUNNING;
883 else
884 ifp->if_flags &= ~IFF_RUNNING;
885 break;
886 case SIOCSIFMTU:
887 if (ifr->ifr_mtu < PFSYNC_MINMTU)
888 return (EINVAL);
889 if (ifr->ifr_mtu > MCLBYTES)
890 ifr->ifr_mtu = MCLBYTES;
891 s = splnet();
892 if (ifr->ifr_mtu < ifp->if_mtu)
893 pfsync_sendout(sc);
894 pfsync_setmtu(sc, ifr->ifr_mtu);
895 splx(s);
896 break;
897 case SIOCGETPFSYNC:
898 if ((error = kauth_authorize_network(l->l_cred,
899 KAUTH_NETWORK_INTERFACE,
900 KAUTH_REQ_NETWORK_INTERFACE_GETPRIV, ifp, (void *)cmd,
901 NULL)) != 0)
902 return (error);
903 memset(&pfsyncr, 0, sizeof(pfsyncr));
904 if (sc->sc_sync_ifp)
905 strlcpy(pfsyncr.pfsyncr_syncdev,
906 sc->sc_sync_ifp->if_xname, IFNAMSIZ);
907 pfsyncr.pfsyncr_syncpeer = sc->sc_sync_peer;
908 pfsyncr.pfsyncr_maxupdates = sc->sc_maxupdates;
909 if ((error = copyout(&pfsyncr, ifr->ifr_data, sizeof(pfsyncr))))
910 return (error);
911 break;
912 case SIOCSETPFSYNC:
913 if ((error = kauth_authorize_network(l->l_cred,
914 KAUTH_NETWORK_INTERFACE,
915 KAUTH_REQ_NETWORK_INTERFACE_SETPRIV, ifp, (void *)cmd,
916 NULL)) != 0)
917 return (error);
918 if ((error = copyin(ifr->ifr_data, &pfsyncr, sizeof(pfsyncr))))
919 return (error);
920
921 if (pfsyncr.pfsyncr_syncpeer.s_addr == 0)
922 sc->sc_sync_peer.s_addr = INADDR_PFSYNC_GROUP;
923 else
924 sc->sc_sync_peer.s_addr =
925 pfsyncr.pfsyncr_syncpeer.s_addr;
926
927 if (pfsyncr.pfsyncr_maxupdates > 255)
928 return (EINVAL);
929 sc->sc_maxupdates = pfsyncr.pfsyncr_maxupdates;
930
931 if (pfsyncr.pfsyncr_syncdev[0] == 0) {
932 sc->sc_sync_ifp = NULL;
933 if (sc->sc_mbuf_net != NULL) {
934 /* Don't keep stale pfsync packets around. */
935 s = splnet();
936 m_freem(sc->sc_mbuf_net);
937 sc->sc_mbuf_net = NULL;
938 sc->sc_statep_net.s = NULL;
939 splx(s);
940 }
941 if (imo->imo_num_memberships > 0) {
942 in_delmulti(imo->imo_membership[--imo->imo_num_memberships]);
943 imo->imo_multicast_if_index = 0;
944 }
945 break;
946 }
947
948 if ((sifp = ifunit(pfsyncr.pfsyncr_syncdev)) == NULL)
949 return (EINVAL);
950
951 s = splnet();
952 if (sifp->if_mtu < sc->sc_if.if_mtu ||
953 (sc->sc_sync_ifp != NULL &&
954 sifp->if_mtu < sc->sc_sync_ifp->if_mtu) ||
955 sifp->if_mtu < MCLBYTES - sizeof(struct ip))
956 pfsync_sendout(sc);
957 sc->sc_sync_ifp = sifp;
958
959 pfsync_setmtu(sc, sc->sc_if.if_mtu);
960
961 if (imo->imo_num_memberships > 0) {
962 in_delmulti(imo->imo_membership[--imo->imo_num_memberships]);
963 imo->imo_multicast_if_index = 0;
964 }
965
966 if (sc->sc_sync_ifp &&
967 sc->sc_sync_peer.s_addr == INADDR_PFSYNC_GROUP) {
968 struct in_addr addr;
969
970 if (!(sc->sc_sync_ifp->if_flags & IFF_MULTICAST)) {
971 sc->sc_sync_ifp = NULL;
972 splx(s);
973 return (EADDRNOTAVAIL);
974 }
975
976 addr.s_addr = INADDR_PFSYNC_GROUP;
977
978 if ((imo->imo_membership[0] =
979 in_addmulti(&addr, sc->sc_sync_ifp)) == NULL) {
980 sc->sc_sync_ifp = NULL;
981 splx(s);
982 return (ENOBUFS);
983 }
984 imo->imo_num_memberships++;
985 imo->imo_multicast_if_index = if_get_index(sc->sc_sync_ifp);
986 imo->imo_multicast_ttl = PFSYNC_DFLTTL;
987 imo->imo_multicast_loop = 0;
988 }
989
990 if (sc->sc_sync_ifp ||
991 sc->sc_sendaddr.s_addr != INADDR_PFSYNC_GROUP) {
992 /* Request a full state table update. */
993 sc->sc_ureq_sent = time_uptime;
994 #if NCARP > 0
995 if (pfsync_sync_ok)
996 carp_suppress_preempt ++;
997 #endif
998 pfsync_sync_ok = 0;
999 if (pf_status.debug >= PF_DEBUG_MISC)
1000 printf("pfsync: requesting bulk update\n");
1001 callout_schedule(&sc->sc_bulkfail_tmo, 5 * hz);
1002 error = pfsync_request_update(NULL, NULL);
1003 if (error == ENOMEM) {
1004 splx(s);
1005 return (ENOMEM);
1006 }
1007 pfsync_sendout(sc);
1008 }
1009 splx(s);
1010
1011 break;
1012
1013 default:
1014 return ifioctl_common(ifp, cmd, data);
1015 }
1016
1017 return (0);
1018 }
1019
1020 void
pfsync_setmtu(struct pfsync_softc * sc,int mtu_req)1021 pfsync_setmtu(struct pfsync_softc *sc, int mtu_req)
1022 {
1023 int mtu;
1024
1025 if (sc->sc_sync_ifp && sc->sc_sync_ifp->if_mtu < mtu_req)
1026 mtu = sc->sc_sync_ifp->if_mtu;
1027 else
1028 mtu = mtu_req;
1029
1030 sc->sc_maxcount = (mtu - sizeof(struct pfsync_header)) /
1031 sizeof(struct pfsync_state);
1032 if (sc->sc_maxcount > 254)
1033 sc->sc_maxcount = 254;
1034 sc->sc_if.if_mtu = sizeof(struct pfsync_header) +
1035 sc->sc_maxcount * sizeof(struct pfsync_state);
1036 }
1037
1038 struct mbuf *
pfsync_get_mbuf(struct pfsync_softc * sc,u_int8_t action,void ** sp)1039 pfsync_get_mbuf(struct pfsync_softc *sc, u_int8_t action, void **sp)
1040 {
1041 struct pfsync_header *h;
1042 struct mbuf *m;
1043 int len;
1044
1045 MGETHDR(m, M_DONTWAIT, MT_DATA);
1046 if (m == NULL) {
1047 #ifdef __NetBSD__
1048 if_statinc(&sc->sc_if, if_oerrors);
1049 #else
1050 sc->sc_if.if_oerrors++;
1051 #endif /* __NetBSD__ */
1052 return (NULL);
1053 }
1054
1055 switch (action) {
1056 case PFSYNC_ACT_CLR:
1057 len = sizeof(struct pfsync_header) +
1058 sizeof(struct pfsync_state_clr);
1059 break;
1060 case PFSYNC_ACT_UPD_C:
1061 len = (sc->sc_maxcount * sizeof(struct pfsync_state_upd)) +
1062 sizeof(struct pfsync_header);
1063 break;
1064 case PFSYNC_ACT_DEL_C:
1065 len = (sc->sc_maxcount * sizeof(struct pfsync_state_del)) +
1066 sizeof(struct pfsync_header);
1067 break;
1068 case PFSYNC_ACT_UREQ:
1069 len = (sc->sc_maxcount * sizeof(struct pfsync_state_upd_req)) +
1070 sizeof(struct pfsync_header);
1071 break;
1072 case PFSYNC_ACT_BUS:
1073 len = sizeof(struct pfsync_header) +
1074 sizeof(struct pfsync_state_bus);
1075 break;
1076 case PFSYNC_ACT_TDB_UPD:
1077 len = (sc->sc_maxcount * sizeof(struct pfsync_tdb)) +
1078 sizeof(struct pfsync_header);
1079 break;
1080 default:
1081 len = (sc->sc_maxcount * sizeof(struct pfsync_state)) +
1082 sizeof(struct pfsync_header);
1083 break;
1084 }
1085
1086 if (len > MHLEN) {
1087 MCLGET(m, M_DONTWAIT);
1088 if ((m->m_flags & M_EXT) == 0) {
1089 m_free(m);
1090 #ifdef __NetBSD__
1091 if_statinc(&sc->sc_if, if_oerrors);
1092 #else
1093 sc->sc_if.if_oerrors++;
1094 #endif /* __NetBSD__ */
1095 return (NULL);
1096 }
1097 m->m_data += (MCLBYTES - len) &~ (sizeof(long) - 1);
1098 } else
1099 m_align(m, len);
1100
1101 m_reset_rcvif(m);
1102 m->m_pkthdr.len = m->m_len = sizeof(struct pfsync_header);
1103 h = mtod(m, struct pfsync_header *);
1104 h->version = PFSYNC_VERSION;
1105 h->af = 0;
1106 h->count = 0;
1107 h->action = action;
1108 if (action != PFSYNC_ACT_TDB_UPD)
1109 memcpy(&h->pf_chksum, &pf_status.pf_chksum,
1110 PF_MD5_DIGEST_LENGTH);
1111
1112 *sp = (void *)((char *)h + PFSYNC_HDRLEN);
1113 if (action == PFSYNC_ACT_TDB_UPD)
1114 callout_schedule(&sc->sc_tdb_tmo, hz);
1115 else
1116 callout_schedule(&sc->sc_tmo, hz);
1117 return (m);
1118 }
1119
1120 int
pfsync_pack_state(u_int8_t action,struct pf_state * st,int flags)1121 pfsync_pack_state(u_int8_t action, struct pf_state *st, int flags)
1122 {
1123 struct ifnet *ifp = NULL;
1124 struct pfsync_softc *sc = pfsyncif;
1125 struct pfsync_header *h, *h_net;
1126 struct pfsync_state *sp = NULL;
1127 struct pfsync_state_upd *up = NULL;
1128 struct pfsync_state_del *dp = NULL;
1129 struct pf_state_key *sk = st->state_key;
1130 struct pf_rule *r;
1131 u_long secs;
1132 int s, ret = 0;
1133 u_int8_t i = 255, newaction = 0;
1134
1135 if (sc == NULL)
1136 return (0);
1137 ifp = &sc->sc_if;
1138
1139 /*
1140 * If a packet falls in the forest and there's nobody around to
1141 * hear, does it make a sound?
1142 */
1143 if (ifp->if_bpf == NULL && sc->sc_sync_ifp == NULL &&
1144 sc->sc_sync_peer.s_addr == INADDR_PFSYNC_GROUP) {
1145 /* Don't leave any stale pfsync packets hanging around. */
1146 if (sc->sc_mbuf != NULL) {
1147 m_freem(sc->sc_mbuf);
1148 sc->sc_mbuf = NULL;
1149 sc->sc_statep.s = NULL;
1150 }
1151 return (0);
1152 }
1153
1154 if (action >= PFSYNC_ACT_MAX)
1155 return (EINVAL);
1156
1157 s = splnet();
1158 if (sc->sc_mbuf == NULL) {
1159 if ((sc->sc_mbuf = pfsync_get_mbuf(sc, action,
1160 (void *)&sc->sc_statep.s)) == NULL) {
1161 splx(s);
1162 return (ENOMEM);
1163 }
1164 h = mtod(sc->sc_mbuf, struct pfsync_header *);
1165 } else {
1166 h = mtod(sc->sc_mbuf, struct pfsync_header *);
1167 if (h->action != action) {
1168 pfsync_sendout(sc);
1169 if ((sc->sc_mbuf = pfsync_get_mbuf(sc, action,
1170 (void *)&sc->sc_statep.s)) == NULL) {
1171 splx(s);
1172 return (ENOMEM);
1173 }
1174 h = mtod(sc->sc_mbuf, struct pfsync_header *);
1175 } else {
1176 /*
1177 * If it's an update, look in the packet to see if
1178 * we already have an update for the state.
1179 */
1180 if (action == PFSYNC_ACT_UPD && sc->sc_maxupdates) {
1181 struct pfsync_state *usp =
1182 (void *)((char *)h + PFSYNC_HDRLEN);
1183
1184 for (i = 0; i < h->count; i++) {
1185 if (!memcmp(usp->id, &st->id,
1186 PFSYNC_ID_LEN) &&
1187 usp->creatorid == st->creatorid) {
1188 sp = usp;
1189 sp->updates++;
1190 break;
1191 }
1192 usp++;
1193 }
1194 }
1195 }
1196 }
1197
1198 secs = time_second;
1199
1200 st->pfsync_time = time_uptime;
1201
1202 if (sp == NULL) {
1203 /* not a "duplicate" update */
1204 i = 255;
1205 sp = sc->sc_statep.s++;
1206 sc->sc_mbuf->m_pkthdr.len =
1207 sc->sc_mbuf->m_len += sizeof(struct pfsync_state);
1208 h->count++;
1209 memset(sp, 0, sizeof(*sp));
1210
1211 memcpy(sp->id, &st->id, sizeof(sp->id));
1212 sp->creatorid = st->creatorid;
1213
1214 strlcpy(sp->ifname, st->kif->pfik_name, sizeof(sp->ifname));
1215 pf_state_host_hton(&sk->lan, &sp->lan);
1216 pf_state_host_hton(&sk->gwy, &sp->gwy);
1217 pf_state_host_hton(&sk->ext, &sp->ext);
1218
1219 memcpy(&sp->rt_addr, &st->rt_addr, sizeof(sp->rt_addr));
1220
1221 sp->creation = htonl(secs - st->creation);
1222 pf_state_counter_hton(st->packets[0], sp->packets[0]);
1223 pf_state_counter_hton(st->packets[1], sp->packets[1]);
1224 pf_state_counter_hton(st->bytes[0], sp->bytes[0]);
1225 pf_state_counter_hton(st->bytes[1], sp->bytes[1]);
1226 if ((r = st->rule.ptr) == NULL)
1227 sp->rule = htonl(-1);
1228 else
1229 sp->rule = htonl(r->nr);
1230 if ((r = st->anchor.ptr) == NULL)
1231 sp->anchor = htonl(-1);
1232 else
1233 sp->anchor = htonl(r->nr);
1234 sp->af = sk->af;
1235 sp->proto = sk->proto;
1236 sp->direction = sk->direction;
1237 sp->log = st->log;
1238 sp->allow_opts = st->allow_opts;
1239 sp->timeout = st->timeout;
1240
1241 if (flags & PFSYNC_FLAG_STALE)
1242 sp->sync_flags |= PFSTATE_STALE;
1243 }
1244
1245 pf_state_peer_hton(&st->src, &sp->src);
1246 pf_state_peer_hton(&st->dst, &sp->dst);
1247
1248 if (st->expire <= secs)
1249 sp->expire = htonl(0);
1250 else
1251 sp->expire = htonl(st->expire - secs);
1252
1253 /* do we need to build "compressed" actions for network transfer? */
1254 if (sc->sc_sync_ifp && flags & PFSYNC_FLAG_COMPRESS) {
1255 switch (action) {
1256 case PFSYNC_ACT_UPD:
1257 newaction = PFSYNC_ACT_UPD_C;
1258 break;
1259 case PFSYNC_ACT_DEL:
1260 newaction = PFSYNC_ACT_DEL_C;
1261 break;
1262 default:
1263 /* by default we just send the uncompressed states */
1264 break;
1265 }
1266 }
1267
1268 if (newaction) {
1269 if (sc->sc_mbuf_net == NULL) {
1270 if ((sc->sc_mbuf_net = pfsync_get_mbuf(sc, newaction,
1271 (void *)&sc->sc_statep_net.s)) == NULL) {
1272 splx(s);
1273 return (ENOMEM);
1274 }
1275 }
1276 h_net = mtod(sc->sc_mbuf_net, struct pfsync_header *);
1277
1278 switch (newaction) {
1279 case PFSYNC_ACT_UPD_C:
1280 if (i != 255) {
1281 up = (void *)((char *)h_net +
1282 PFSYNC_HDRLEN + (i * sizeof(*up)));
1283 up->updates++;
1284 } else {
1285 h_net->count++;
1286 sc->sc_mbuf_net->m_pkthdr.len =
1287 sc->sc_mbuf_net->m_len += sizeof(*up);
1288 up = sc->sc_statep_net.u++;
1289
1290 memset(up, 0, sizeof(*up));
1291 memcpy(up->id, &st->id, sizeof(up->id));
1292 up->creatorid = st->creatorid;
1293 }
1294 up->timeout = st->timeout;
1295 up->expire = sp->expire;
1296 up->src = sp->src;
1297 up->dst = sp->dst;
1298 break;
1299 case PFSYNC_ACT_DEL_C:
1300 sc->sc_mbuf_net->m_pkthdr.len =
1301 sc->sc_mbuf_net->m_len += sizeof(*dp);
1302 dp = sc->sc_statep_net.d++;
1303 h_net->count++;
1304
1305 memset(dp, 0, sizeof(*dp));
1306 memcpy(dp->id, &st->id, sizeof(dp->id));
1307 dp->creatorid = st->creatorid;
1308 break;
1309 }
1310 }
1311
1312 if (h->count == sc->sc_maxcount ||
1313 (sc->sc_maxupdates && (sp->updates >= sc->sc_maxupdates)))
1314 ret = pfsync_sendout(sc);
1315
1316 splx(s);
1317 return (ret);
1318 }
1319
1320 /* This must be called in splnet() */
1321 int
pfsync_request_update(struct pfsync_state_upd * up,struct in_addr * src)1322 pfsync_request_update(struct pfsync_state_upd *up, struct in_addr *src)
1323 {
1324 struct pfsync_header *h;
1325 struct pfsync_softc *sc = pfsyncif;
1326 struct pfsync_state_upd_req *rup;
1327 int ret = 0;
1328
1329 if (sc == NULL)
1330 return (0);
1331
1332 if (sc->sc_mbuf == NULL) {
1333 if ((sc->sc_mbuf = pfsync_get_mbuf(sc, PFSYNC_ACT_UREQ,
1334 (void *)&sc->sc_statep.s)) == NULL)
1335 return (ENOMEM);
1336 h = mtod(sc->sc_mbuf, struct pfsync_header *);
1337 } else {
1338 h = mtod(sc->sc_mbuf, struct pfsync_header *);
1339 if (h->action != PFSYNC_ACT_UREQ) {
1340 pfsync_sendout(sc);
1341 if ((sc->sc_mbuf = pfsync_get_mbuf(sc, PFSYNC_ACT_UREQ,
1342 (void *)&sc->sc_statep.s)) == NULL)
1343 return (ENOMEM);
1344 h = mtod(sc->sc_mbuf, struct pfsync_header *);
1345 }
1346 }
1347
1348 if (src != NULL)
1349 sc->sc_sendaddr = *src;
1350 sc->sc_mbuf->m_pkthdr.len = sc->sc_mbuf->m_len += sizeof(*rup);
1351 h->count++;
1352 rup = sc->sc_statep.r++;
1353 memset(rup, 0, sizeof(*rup));
1354 if (up != NULL) {
1355 memcpy(rup->id, up->id, sizeof(rup->id));
1356 rup->creatorid = up->creatorid;
1357 }
1358
1359 if (h->count == sc->sc_maxcount)
1360 ret = pfsync_sendout(sc);
1361
1362 return (ret);
1363 }
1364
1365 int
pfsync_clear_states(u_int32_t creatorid,char * ifname)1366 pfsync_clear_states(u_int32_t creatorid, char *ifname)
1367 {
1368 struct pfsync_softc *sc = pfsyncif;
1369 struct pfsync_state_clr *cp;
1370 int s, ret;
1371
1372 if (sc == NULL)
1373 return (0);
1374
1375 s = splnet();
1376 if (sc->sc_mbuf != NULL)
1377 pfsync_sendout(sc);
1378 if ((sc->sc_mbuf = pfsync_get_mbuf(sc, PFSYNC_ACT_CLR,
1379 (void *)&sc->sc_statep.c)) == NULL) {
1380 splx(s);
1381 return (ENOMEM);
1382 }
1383 sc->sc_mbuf->m_pkthdr.len = sc->sc_mbuf->m_len += sizeof(*cp);
1384 cp = sc->sc_statep.c;
1385 cp->creatorid = creatorid;
1386 if (ifname != NULL)
1387 strlcpy(cp->ifname, ifname, IFNAMSIZ);
1388
1389 ret = (pfsync_sendout(sc));
1390 splx(s);
1391 return (ret);
1392 }
1393
1394 void
pfsync_timeout(void * v)1395 pfsync_timeout(void *v)
1396 {
1397 struct pfsync_softc *sc = v;
1398 int s;
1399
1400 s = splnet();
1401 pfsync_sendout(sc);
1402 splx(s);
1403 }
1404
1405 void
pfsync_tdb_timeout(void * v)1406 pfsync_tdb_timeout(void *v)
1407 {
1408 struct pfsync_softc *sc = v;
1409 int s;
1410
1411 s = splnet();
1412 pfsync_tdb_sendout(sc);
1413 splx(s);
1414 }
1415
1416 /* This must be called in splnet() */
1417 void
pfsync_send_bus(struct pfsync_softc * sc,u_int8_t status)1418 pfsync_send_bus(struct pfsync_softc *sc, u_int8_t status)
1419 {
1420 struct pfsync_state_bus *bus;
1421
1422 if (sc->sc_mbuf != NULL)
1423 pfsync_sendout(sc);
1424
1425 if (pfsync_sync_ok &&
1426 (sc->sc_mbuf = pfsync_get_mbuf(sc, PFSYNC_ACT_BUS,
1427 (void *)&sc->sc_statep.b)) != NULL) {
1428 sc->sc_mbuf->m_pkthdr.len = sc->sc_mbuf->m_len += sizeof(*bus);
1429 bus = sc->sc_statep.b;
1430 bus->creatorid = pf_status.hostid;
1431 bus->status = status;
1432 bus->endtime = htonl(time_uptime - sc->sc_ureq_received);
1433 pfsync_sendout(sc);
1434 }
1435 }
1436
1437 void
pfsync_bulk_update(void * v)1438 pfsync_bulk_update(void *v)
1439 {
1440 struct pfsync_softc *sc = v;
1441 int s, i = 0;
1442 struct pf_state *state;
1443
1444 s = splnet();
1445 if (sc->sc_mbuf != NULL)
1446 pfsync_sendout(sc);
1447
1448 /*
1449 * Grab at most PFSYNC_BULKPACKETS worth of states which have not
1450 * been sent since the latest request was made.
1451 */
1452 state = sc->sc_bulk_send_next;
1453 if (state)
1454 do {
1455 /* send state update if syncable and not already sent */
1456 if (!state->sync_flags
1457 && state->timeout < PFTM_MAX
1458 && state->pfsync_time <= sc->sc_ureq_received) {
1459 pfsync_pack_state(PFSYNC_ACT_UPD, state, 0);
1460 i++;
1461 }
1462
1463 /* figure next state to send */
1464 state = TAILQ_NEXT(state, entry_list);
1465
1466 /* wrap to start of list if we hit the end */
1467 if (!state)
1468 state = TAILQ_FIRST(&state_list);
1469 } while (i < sc->sc_maxcount * PFSYNC_BULKPACKETS &&
1470 state != sc->sc_bulk_terminator);
1471
1472 if (!state || state == sc->sc_bulk_terminator) {
1473 /* we're done */
1474 pfsync_send_bus(sc, PFSYNC_BUS_END);
1475 sc->sc_ureq_received = 0;
1476 sc->sc_bulk_send_next = NULL;
1477 sc->sc_bulk_terminator = NULL;
1478 callout_stop(&sc->sc_bulk_tmo);
1479 if (pf_status.debug >= PF_DEBUG_MISC)
1480 printf("pfsync: bulk update complete\n");
1481 } else {
1482 /* look again for more in a bit */
1483 callout_schedule(&sc->sc_bulk_tmo, 1);
1484 sc->sc_bulk_send_next = state;
1485 }
1486 if (sc->sc_mbuf != NULL)
1487 pfsync_sendout(sc);
1488 splx(s);
1489 }
1490
1491 void
pfsync_bulkfail(void * v)1492 pfsync_bulkfail(void *v)
1493 {
1494 struct pfsync_softc *sc = v;
1495 int s, error;
1496
1497 if (sc->sc_bulk_tries++ < PFSYNC_MAX_BULKTRIES) {
1498 /* Try again in a bit */
1499 callout_schedule(&sc->sc_bulkfail_tmo, 5 * hz);
1500 s = splnet();
1501 error = pfsync_request_update(NULL, NULL);
1502 if (error == ENOMEM) {
1503 if (pf_status.debug >= PF_DEBUG_MISC)
1504 printf("pfsync: cannot allocate mbufs for "
1505 "bulk update\n");
1506 } else
1507 pfsync_sendout(sc);
1508 splx(s);
1509 } else {
1510 /* Pretend like the transfer was ok */
1511 sc->sc_ureq_sent = 0;
1512 sc->sc_bulk_tries = 0;
1513 #if NCARP > 0
1514 if (!pfsync_sync_ok)
1515 carp_suppress_preempt --;
1516 #endif
1517 pfsync_sync_ok = 1;
1518 if (pf_status.debug >= PF_DEBUG_MISC)
1519 printf("pfsync: failed to receive "
1520 "bulk update status\n");
1521 callout_stop(&sc->sc_bulkfail_tmo);
1522 }
1523 }
1524
1525 /* This must be called in splnet() */
1526 int
pfsync_sendout(struct pfsync_softc * sc)1527 pfsync_sendout(struct pfsync_softc *sc)
1528 {
1529 struct ifnet *ifp = &sc->sc_if;
1530 struct mbuf *m;
1531
1532 callout_stop(&sc->sc_tmo);
1533
1534 if (sc->sc_mbuf == NULL)
1535 return (0);
1536 m = sc->sc_mbuf;
1537 sc->sc_mbuf = NULL;
1538 sc->sc_statep.s = NULL;
1539
1540 bpf_mtap(ifp, m, BPF_D_OUT);
1541
1542 if (sc->sc_mbuf_net) {
1543 m_freem(m);
1544 m = sc->sc_mbuf_net;
1545 sc->sc_mbuf_net = NULL;
1546 sc->sc_statep_net.s = NULL;
1547 }
1548
1549 return pfsync_sendout_mbuf(sc, m);
1550 }
1551
1552 int
pfsync_tdb_sendout(struct pfsync_softc * sc)1553 pfsync_tdb_sendout(struct pfsync_softc *sc)
1554 {
1555 struct ifnet *ifp = &sc->sc_if;
1556 struct mbuf *m;
1557
1558 callout_stop(&sc->sc_tdb_tmo);
1559
1560 if (sc->sc_mbuf_tdb == NULL)
1561 return (0);
1562 m = sc->sc_mbuf_tdb;
1563 sc->sc_mbuf_tdb = NULL;
1564 sc->sc_statep_tdb.t = NULL;
1565
1566 bpf_mtap(ifp, m, BPF_D_OUT);
1567
1568 return pfsync_sendout_mbuf(sc, m);
1569 }
1570
1571 int
pfsync_sendout_mbuf(struct pfsync_softc * sc,struct mbuf * m)1572 pfsync_sendout_mbuf(struct pfsync_softc *sc, struct mbuf *m)
1573 {
1574 struct sockaddr sa;
1575 struct ip *ip;
1576
1577 if (sc->sc_sync_ifp ||
1578 sc->sc_sync_peer.s_addr != INADDR_PFSYNC_GROUP) {
1579 M_PREPEND(m, sizeof(struct ip), M_DONTWAIT);
1580 if (m == NULL) {
1581 PFSYNC_STATINC(PFSYNC_STAT_ONOMEM);
1582 return (0);
1583 }
1584 ip = mtod(m, struct ip *);
1585 ip->ip_v = IPVERSION;
1586 ip->ip_hl = sizeof(*ip) >> 2;
1587 ip->ip_tos = IPTOS_LOWDELAY;
1588 ip->ip_len = htons(m->m_pkthdr.len);
1589 ip->ip_id = ip_randomid();
1590 ip->ip_off = htons(IP_DF);
1591 ip->ip_ttl = PFSYNC_DFLTTL;
1592 ip->ip_p = IPPROTO_PFSYNC;
1593 ip->ip_sum = 0;
1594
1595 memset(&sa, 0, sizeof(sa));
1596 ip->ip_src.s_addr = INADDR_ANY;
1597
1598 if (sc->sc_sendaddr.s_addr == INADDR_PFSYNC_GROUP)
1599 m->m_flags |= M_MCAST;
1600 ip->ip_dst = sc->sc_sendaddr;
1601 sc->sc_sendaddr.s_addr = sc->sc_sync_peer.s_addr;
1602
1603 PFSYNC_STATINC(PFSYNC_STAT_OPACKETS);
1604
1605 if (ip_output(m, NULL, NULL, IP_RAWOUTPUT, &sc->sc_imo, NULL)) {
1606 PFSYNC_STATINC(PFSYNC_STAT_OERRORS);
1607 }
1608 } else
1609 m_freem(m);
1610
1611 return (0);
1612 }
1613
1614 static int
sysctl_net_inet_pfsync_stats(SYSCTLFN_ARGS)1615 sysctl_net_inet_pfsync_stats(SYSCTLFN_ARGS)
1616 {
1617
1618 return (NETSTAT_SYSCTL(pfsyncstat_percpu, PFSYNC_NSTATS));
1619 }
1620
1621 SYSCTL_SETUP(sysctl_net_inet_pfsync_setup, "sysctl net.inet.pfsync subtree setup")
1622 {
1623
1624 sysctl_createv(clog, 0, NULL, NULL,
1625 CTLFLAG_PERMANENT,
1626 CTLTYPE_NODE, "net", NULL,
1627 NULL, 0, NULL, 0,
1628 CTL_NET, CTL_EOL);
1629 sysctl_createv(clog, 0, NULL, NULL,
1630 CTLFLAG_PERMANENT,
1631 CTLTYPE_NODE, "inet", NULL,
1632 NULL, 0, NULL, 0,
1633 CTL_NET, PF_INET, CTL_EOL);
1634 sysctl_createv(clog, 0, NULL, NULL,
1635 CTLFLAG_PERMANENT,
1636 CTLTYPE_NODE, "pfsync",
1637 SYSCTL_DESCR("pfsync related settings"),
1638 NULL, 0, NULL, 0,
1639 CTL_NET, PF_INET, IPPROTO_PFSYNC, CTL_EOL);
1640 sysctl_createv(clog, 0, NULL, NULL,
1641 CTLFLAG_PERMANENT|CTLFLAG_READONLY,
1642 CTLTYPE_STRUCT, "stats",
1643 SYSCTL_DESCR("pfsync statistics"),
1644 sysctl_net_inet_pfsync_stats, 0, NULL, 0,
1645 CTL_NET, PF_INET, IPPROTO_PFSYNC,
1646 CTL_CREATE, CTL_EOL);
1647 }
1648