Adapt OpenBSD pf's "sloopy" TCP state machine which is useful for Direct

Server Return mode, where not all packets would be visible to the load
balancer or gateway.

This commit should be reverted when we merge future pf versions.  The
benefit it would provide is that this version does not break any existing
public interface and thus won't be a problem if we want to MFC it to
earlier FreeBSD releases.

Discussed with:	mlaier
Obtained from:	OpenBSD
Sponsored by:	iXsystems, Inc.
MFC after:	1 month

4 files changed, 296 insertions, 175 deletions

diff --git a/sys/contrib/pf/net/if_pfsync.c b/sys/contrib/pf/net/if_pfsync.c
index 0925edb094b9..946611859175 100644
--- a/sys/contrib/pf/net/if_pfsync.c
+++ b/sys/contrib/pf/net/if_pfsync.c
@@ -465,7 +465,7 @@ pfsync_insert_net_state(struct pfsync_state *sp, u_int8_t chksum_flag)
 	st->direction = sp->direction;
 	st->log = sp->log;
 	st->timeout = sp->timeout;
-	st->allow_opts = sp->allow_opts;
+	st->state_flags = sp->state_flags;
 
 	bcopy(sp->id, &st->id, sizeof(st->id));
 	st->creatorid = sp->creatorid;
@@ -1578,7 +1578,7 @@ pfsync_pack_state(u_int8_t action, struct pf_state *st, int flags)
 		sp->proto = st->proto;
 		sp->direction = st->direction;
 		sp->log = st->log;
-		sp->allow_opts = st->allow_opts;
+		sp->state_flags = st->state_flags;
 		sp->timeout = st->timeout;
 
 		if (flags & PFSYNC_FLAG_STALE)
diff --git a/sys/contrib/pf/net/if_pfsync.h b/sys/contrib/pf/net/if_pfsync.h
index f7edca3e4003..f306610fb3d5 100644
--- a/sys/contrib/pf/net/if_pfsync.h
+++ b/sys/contrib/pf/net/if_pfsync.h
@@ -80,7 +80,7 @@ struct pfsync_state {
 	u_int8_t	 proto;
 	u_int8_t	 direction;
 	u_int8_t	 log;
-	u_int8_t	 allow_opts;
+	u_int8_t	 state_flags;
 	u_int8_t	 timeout;
 	u_int8_t	 sync_flags;
 	u_int8_t	 updates;
diff --git a/sys/contrib/pf/net/pf.c b/sys/contrib/pf/net/pf.c
index e8031ac7dfec..ebe73c2e9320 100644
--- a/sys/contrib/pf/net/pf.c
+++ b/sys/contrib/pf/net/pf.c
@@ -253,6 +253,13 @@ int			 pf_test_fragment(struct pf_rule **, int,
 			    struct pfi_kif *, struct mbuf *, void *,
 			    struct pf_pdesc *, struct pf_rule **,
 			    struct pf_ruleset **);
+int			 pf_tcp_track_full(struct pf_state_peer *,
+			    struct pf_state_peer *, struct pf_state **,
+			    struct pfi_kif *, struct mbuf *, int,
+			    struct pf_pdesc *, u_short *, int *);
+int			 pf_tcp_track_sloppy(struct pf_state_peer *,
+			    struct pf_state_peer *, struct pf_state **,
+			    struct pf_pdesc *, u_short *);
 int			 pf_test_state_tcp(struct pf_state **, int,
 			    struct pfi_kif *, struct mbuf *, int,
 			    void *, struct pf_pdesc *, u_short *);
@@ -3528,7 +3535,10 @@ cleanup:
 		s->nat_rule.ptr = nr;
 		s->anchor.ptr = a;
 		STATE_INC_COUNTERS(s);
-		s->allow_opts = r->allow_opts;
+		if (r->allow_opts)
+			s->state_flags |= PFSTATE_ALLOWOPTS;
+		if (r->rule_flag & PFRULE_STATESLOPPY)
+			s->state_flags |= PFSTATE_SLOPPY;
 		s->log = r->log & PF_LOG_ALL;
 		if (nr != NULL)
 			s->log |= nr->log & PF_LOG_ALL;
@@ -3925,7 +3935,10 @@ cleanup:
 		s->nat_rule.ptr = nr;
 		s->anchor.ptr = a;
 		STATE_INC_COUNTERS(s);
-		s->allow_opts = r->allow_opts;
+		if (r->allow_opts)
+			s->state_flags |= PFSTATE_ALLOWOPTS;
+ 		if (r->rule_flag & PFRULE_STATESLOPPY)
+			s->state_flags |= PFSTATE_SLOPPY;
 		s->log = r->log & PF_LOG_ALL;
 		if (nr != NULL)
 			s->log |= nr->log & PF_LOG_ALL;
@@ -4238,7 +4251,10 @@ cleanup:
 		s->nat_rule.ptr = nr;
 		s->anchor.ptr = a;
 		STATE_INC_COUNTERS(s);
-		s->allow_opts = r->allow_opts;
+		if (r->allow_opts)
+			s->state_flags |= PFSTATE_ALLOWOPTS;
+ 		if (r->rule_flag & PFRULE_STATESLOPPY)
+			s->state_flags |= PFSTATE_SLOPPY;
 		s->log = r->log & PF_LOG_ALL;
 		if (nr != NULL)
 			s->log |= nr->log & PF_LOG_ALL;
@@ -4525,7 +4541,10 @@ cleanup:
 		s->nat_rule.ptr = nr;
 		s->anchor.ptr = a;
 		STATE_INC_COUNTERS(s);
-		s->allow_opts = r->allow_opts;
+		if (r->allow_opts)
+			s->state_flags |= PFSTATE_ALLOWOPTS;
+ 		if (r->rule_flag & PFRULE_STATESLOPPY)
+			s->state_flags |= PFSTATE_SLOPPY;
 		s->log = r->log & PF_LOG_ALL;
 		if (nr != NULL)
 			s->log |= nr->log & PF_LOG_ALL;
@@ -4666,165 +4685,15 @@ pf_test_fragment(struct pf_rule **rm, int direction, struct pfi_kif *kif,
 }
 
 int
-pf_test_state_tcp(struct pf_state **state, int direction, struct pfi_kif *kif,
-    struct mbuf *m, int off, void *h, struct pf_pdesc *pd,
-    u_short *reason)
+pf_tcp_track_full(struct pf_state_peer *src, struct pf_state_peer *dst,
+	struct pf_state **state, struct pfi_kif *kif, struct mbuf *m, int off,
+	struct pf_pdesc *pd, u_short *reason, int *copyback)
 {
-	struct pf_state_cmp	 key;
-	struct tcphdr		*th = pd->hdr.tcp;
-	u_int16_t		 win = ntohs(th->th_win);
-	u_int32_t		 ack, end, seq, orig_seq;
-	u_int8_t		 sws, dws;
-	int			 ackskew;
-	int			 copyback = 0;
-	struct pf_state_peer	*src, *dst;
-
-	key.af = pd->af;
-	key.proto = IPPROTO_TCP;
-	if (direction == PF_IN)	{
-		PF_ACPY(&key.ext.addr, pd->src, key.af);
-		PF_ACPY(&key.gwy.addr, pd->dst, key.af);
-		key.ext.port = th->th_sport;
-		key.gwy.port = th->th_dport;
-	} else {
-		PF_ACPY(&key.lan.addr, pd->src, key.af);
-		PF_ACPY(&key.ext.addr, pd->dst, key.af);
-		key.lan.port = th->th_sport;
-		key.ext.port = th->th_dport;
-	}
-
-	STATE_LOOKUP();
-
-	if (direction == (*state)->direction) {
-		src = &(*state)->src;
-		dst = &(*state)->dst;
-	} else {
-		src = &(*state)->dst;
-		dst = &(*state)->src;
-	}
-
-	if ((*state)->src.state == PF_TCPS_PROXY_SRC) {
-		if (direction != (*state)->direction) {
-			REASON_SET(reason, PFRES_SYNPROXY);
-			return (PF_SYNPROXY_DROP);
-		}
-		if (th->th_flags & TH_SYN) {
-			if (ntohl(th->th_seq) != (*state)->src.seqlo) {
-				REASON_SET(reason, PFRES_SYNPROXY);
-				return (PF_DROP);
-			}
-#ifdef __FreeBSD__
-			pf_send_tcp(NULL, (*state)->rule.ptr, pd->af, pd->dst,
-#else
-			pf_send_tcp((*state)->rule.ptr, pd->af, pd->dst,
-#endif
-			    pd->src, th->th_dport, th->th_sport,
-			    (*state)->src.seqhi, ntohl(th->th_seq) + 1,
-			    TH_SYN|TH_ACK, 0, (*state)->src.mss, 0, 1,
-			    0, NULL, NULL);
-			REASON_SET(reason, PFRES_SYNPROXY);
-			return (PF_SYNPROXY_DROP);
-		} else if (!(th->th_flags & TH_ACK) ||
-		    (ntohl(th->th_ack) != (*state)->src.seqhi + 1) ||
-		    (ntohl(th->th_seq) != (*state)->src.seqlo + 1)) {
-			REASON_SET(reason, PFRES_SYNPROXY);
-			return (PF_DROP);
-		} else if ((*state)->src_node != NULL &&
-		    pf_src_connlimit(state)) {
-			REASON_SET(reason, PFRES_SRCLIMIT);
-			return (PF_DROP);
-		} else
-			(*state)->src.state = PF_TCPS_PROXY_DST;
-	}
-	if ((*state)->src.state == PF_TCPS_PROXY_DST) {
-		struct pf_state_host *src, *dst;
-
-		if (direction == PF_OUT) {
-			src = &(*state)->gwy;
-			dst = &(*state)->ext;
-		} else {
-			src = &(*state)->ext;
-			dst = &(*state)->lan;
-		}
-		if (direction == (*state)->direction) {
-			if (((th->th_flags & (TH_SYN|TH_ACK)) != TH_ACK) ||
-			    (ntohl(th->th_ack) != (*state)->src.seqhi + 1) ||
-			    (ntohl(th->th_seq) != (*state)->src.seqlo + 1)) {
-				REASON_SET(reason, PFRES_SYNPROXY);
-				return (PF_DROP);
-			}
-			(*state)->src.max_win = MAX(ntohs(th->th_win), 1);
-			if ((*state)->dst.seqhi == 1)
-				(*state)->dst.seqhi = htonl(arc4random());
-#ifdef __FreeBSD__
-			pf_send_tcp(NULL, (*state)->rule.ptr, pd->af,
-			    &src->addr,
-#else
-			pf_send_tcp((*state)->rule.ptr, pd->af, &src->addr,
-#endif
-			    &dst->addr, src->port, dst->port,
-			    (*state)->dst.seqhi, 0, TH_SYN, 0,
-			    (*state)->src.mss, 0, 0, (*state)->tag, NULL, NULL);
-			REASON_SET(reason, PFRES_SYNPROXY);
-			return (PF_SYNPROXY_DROP);
-		} else if (((th->th_flags & (TH_SYN|TH_ACK)) !=
-		    (TH_SYN|TH_ACK)) ||
-		    (ntohl(th->th_ack) != (*state)->dst.seqhi + 1)) {
-			REASON_SET(reason, PFRES_SYNPROXY);
-			return (PF_DROP);
-		} else {
-			(*state)->dst.max_win = MAX(ntohs(th->th_win), 1);
-			(*state)->dst.seqlo = ntohl(th->th_seq);
-#ifdef __FreeBSD__
-			pf_send_tcp(NULL, (*state)->rule.ptr, pd->af, pd->dst,
-#else
-			pf_send_tcp((*state)->rule.ptr, pd->af, pd->dst,
-#endif
-			    pd->src, th->th_dport, th->th_sport,
-			    ntohl(th->th_ack), ntohl(th->th_seq) + 1,
-			    TH_ACK, (*state)->src.max_win, 0, 0, 0,
-			    (*state)->tag, NULL, NULL);
-#ifdef __FreeBSD__
-			pf_send_tcp(NULL, (*state)->rule.ptr, pd->af,
-			    &src->addr,
-#else
-			pf_send_tcp((*state)->rule.ptr, pd->af, &src->addr,
-#endif
-			    &dst->addr, src->port, dst->port,
-			    (*state)->src.seqhi + 1, (*state)->src.seqlo + 1,
-			    TH_ACK, (*state)->dst.max_win, 0, 0, 1,
-			    0, NULL, NULL);
-			(*state)->src.seqdiff = (*state)->dst.seqhi -
-			    (*state)->src.seqlo;
-			(*state)->dst.seqdiff = (*state)->src.seqhi -
-			    (*state)->dst.seqlo;
-			(*state)->src.seqhi = (*state)->src.seqlo +
-			    (*state)->dst.max_win;
-			(*state)->dst.seqhi = (*state)->dst.seqlo +
-			    (*state)->src.max_win;
-			(*state)->src.wscale = (*state)->dst.wscale = 0;
-			(*state)->src.state = (*state)->dst.state =
-			    TCPS_ESTABLISHED;
-			REASON_SET(reason, PFRES_SYNPROXY);
-			return (PF_SYNPROXY_DROP);
-		}
-	}
-
-	if (((th->th_flags & (TH_SYN|TH_ACK)) == TH_SYN) &&
-	    dst->state >= TCPS_FIN_WAIT_2 &&
-	    src->state >= TCPS_FIN_WAIT_2) {
-		if (pf_status.debug >= PF_DEBUG_MISC) {
-			printf("pf: state reuse ");
-			pf_print_state(*state);
-			pf_print_flags(th->th_flags);
-			printf("\n");
-		}
-		/* XXX make sure it's the same direction ?? */
-		(*state)->src.state = (*state)->dst.state = TCPS_CLOSED;
-		pf_unlink_state(*state);
-		*state = NULL;
-		return (PF_DROP);
-	}
+ 	struct tcphdr		*th = pd->hdr.tcp;
+ 	u_int16_t		 win = ntohs(th->th_win);
+ 	u_int32_t		 ack, end, seq, orig_seq;
+ 	u_int8_t		 sws, dws;
+ 	int			 ackskew;
 
 	if (src->wscale && dst->wscale && !(th->th_flags & TH_SYN)) {
 		sws = src->wscale & PF_WSCALE_MASK;
@@ -4863,7 +4732,7 @@ pf_test_state_tcp(struct pf_state **state, int direction, struct pfi_kif *kif,
 			pf_change_a(&th->th_seq, &th->th_sum, htonl(seq +
 			    src->seqdiff), 0);
 			pf_change_a(&th->th_ack, &th->th_sum, htonl(ack), 0);
-			copyback = 1;
+			*copyback = 1;
 		} else {
 			ack = ntohl(th->th_ack);
 		}
@@ -4915,7 +4784,7 @@ pf_test_state_tcp(struct pf_state **state, int direction, struct pfi_kif *kif,
 			pf_change_a(&th->th_seq, &th->th_sum, htonl(seq +
 			    src->seqdiff), 0);
 			pf_change_a(&th->th_ack, &th->th_sum, htonl(ack), 0);
-			copyback = 1;
+			*copyback = 1;
 		}
 		end = seq + pd->p_len;
 		if (th->th_flags & TH_SYN)
@@ -4961,7 +4830,7 @@ pf_test_state_tcp(struct pf_state **state, int direction, struct pfi_kif *kif,
 	 */
 	if (dst->seqdiff && (th->th_off << 2) > sizeof(struct tcphdr)) {
 		if (pf_modulate_sack(m, off, pd, th, dst))
-			copyback = 1;
+			*copyback = 1;
 	}
 
 
@@ -4980,7 +4849,7 @@ pf_test_state_tcp(struct pf_state **state, int direction, struct pfi_kif *kif,
 
 		if (dst->scrub || src->scrub) {
 			if (pf_normalize_tcp_stateful(m, off, pd, reason, th,
-			    *state, src, dst, &copyback))
+			    *state, src, dst, copyback))
 				return (PF_DROP);
 		}
 
@@ -5082,7 +4951,7 @@ pf_test_state_tcp(struct pf_state **state, int direction, struct pfi_kif *kif,
 
 		if (dst->scrub || src->scrub) {
 			if (pf_normalize_tcp_stateful(m, off, pd, reason, th,
-			    *state, src, dst, &copyback))
+			    *state, src, dst, copyback))
 				return (PF_DROP);
 		}
 
@@ -5132,7 +5001,11 @@ pf_test_state_tcp(struct pf_state **state, int direction, struct pfi_kif *kif,
 			pf_print_state(*state);
 			pf_print_flags(th->th_flags);
 			printf(" seq=%u (%u) ack=%u len=%u ackskew=%d "
+#ifdef notyet
 			    "pkts=%llu:%llu dir=%s,%s\n",
+#else
+			    "pkts=%llu:%llu%s\n",
+#endif
 			    seq, orig_seq, ack, pd->p_len, ackskew,
 #ifdef __FreeBSD__
 			    (unsigned long long)(*state)->packets[0],
@@ -5140,8 +5013,12 @@ pf_test_state_tcp(struct pf_state **state, int direction, struct pfi_kif *kif,
 #else
 			    (*state)->packets[0], (*state)->packets[1],
 #endif
+#ifdef notyet
 			    direction == PF_IN ? "in" : "out",
 			    direction == (*state)->direction ? "fwd" : "rev");
+#else
+			    "");
+#endif
 			printf("pf: State failure on: %c %c %c %c | %c %c\n",
 			    SEQ_GEQ(src->seqhi, end) ? ' ' : '1',
 			    SEQ_GEQ(seq, src->seqlo - (dst->max_win << dws)) ?
@@ -5156,6 +5033,246 @@ pf_test_state_tcp(struct pf_state **state, int direction, struct pfi_kif *kif,
 	}
 
 	/* Any packets which have gotten here are to be passed */
+	return (PF_PASS);
+}
+
+int
+pf_tcp_track_sloppy(struct pf_state_peer *src, struct pf_state_peer *dst,
+	struct pf_state **state, struct pf_pdesc *pd, u_short *reason)
+{
+	struct tcphdr		*th = pd->hdr.tcp;
+
+	if (th->th_flags & TH_SYN)
+		if (src->state < TCPS_SYN_SENT)
+			src->state = TCPS_SYN_SENT;
+	if (th->th_flags & TH_FIN)
+		if (src->state < TCPS_CLOSING)
+			src->state = TCPS_CLOSING;
+	if (th->th_flags & TH_ACK) {
+		if (dst->state == TCPS_SYN_SENT) {
+			dst->state = TCPS_ESTABLISHED;
+			if (src->state == TCPS_ESTABLISHED &&
+			    (*state)->src_node != NULL &&
+			    pf_src_connlimit(state)) {
+				REASON_SET(reason, PFRES_SRCLIMIT);
+				return (PF_DROP);
+			}
+		} else if (dst->state == TCPS_CLOSING) {
+			dst->state = TCPS_FIN_WAIT_2;
+		} else if (src->state == TCPS_SYN_SENT &&
+		    dst->state < TCPS_SYN_SENT) {
+			/*
+			 * Handle a special sloppy case where we only see one
+			 * half of the connection. If there is a ACK after
+			 * the initial SYN without ever seeing a packet from
+			 * the destination, set the connection to established.
+			 */
+			dst->state = src->state = TCPS_ESTABLISHED;
+			if ((*state)->src_node != NULL &&
+			    pf_src_connlimit(state)) {
+				REASON_SET(reason, PFRES_SRCLIMIT);
+				return (PF_DROP);
+			}
+		} else if (src->state == TCPS_CLOSING &&
+		    dst->state == TCPS_ESTABLISHED &&
+		    dst->seqlo == 0) {
+			/*
+			 * Handle the closing of half connections where we
+			 * don't see the full bidirectional FIN/ACK+ACK
+			 * handshake.
+			 */
+			dst->state = TCPS_CLOSING;
+		}
+	}
+	if (th->th_flags & TH_RST)
+		src->state = dst->state = TCPS_TIME_WAIT;
+
+	/* update expire time */
+	(*state)->expire = time_second;
+	if (src->state >= TCPS_FIN_WAIT_2 &&
+	    dst->state >= TCPS_FIN_WAIT_2)
+		(*state)->timeout = PFTM_TCP_CLOSED;
+	else if (src->state >= TCPS_CLOSING &&
+	    dst->state >= TCPS_CLOSING)
+		(*state)->timeout = PFTM_TCP_FIN_WAIT;
+	else if (src->state < TCPS_ESTABLISHED ||
+	    dst->state < TCPS_ESTABLISHED)
+		(*state)->timeout = PFTM_TCP_OPENING;
+	else if (src->state >= TCPS_CLOSING ||
+	    dst->state >= TCPS_CLOSING)
+		(*state)->timeout = PFTM_TCP_CLOSING;
+	else
+		(*state)->timeout = PFTM_TCP_ESTABLISHED;
+
+	return (PF_PASS);
+}
+
+
+int
+pf_test_state_tcp(struct pf_state **state, int direction, struct pfi_kif *kif,
+    struct mbuf *m, int off, void *h, struct pf_pdesc *pd,
+    u_short *reason)
+{
+	struct pf_state_cmp	 key;
+	struct tcphdr		*th = pd->hdr.tcp;
+	int			 copyback = 0;
+	struct pf_state_peer	*src, *dst;
+
+	key.af = pd->af;
+	key.proto = IPPROTO_TCP;
+	if (direction == PF_IN)	{
+		PF_ACPY(&key.ext.addr, pd->src, key.af);
+		PF_ACPY(&key.gwy.addr, pd->dst, key.af);
+		key.ext.port = th->th_sport;
+		key.gwy.port = th->th_dport;
+	} else {
+		PF_ACPY(&key.lan.addr, pd->src, key.af);
+		PF_ACPY(&key.ext.addr, pd->dst, key.af);
+		key.lan.port = th->th_sport;
+		key.ext.port = th->th_dport;
+	}
+
+	STATE_LOOKUP();
+
+	if (direction == (*state)->direction) {
+		src = &(*state)->src;
+		dst = &(*state)->dst;
+	} else {
+		src = &(*state)->dst;
+		dst = &(*state)->src;
+	}
+
+	if ((*state)->src.state == PF_TCPS_PROXY_SRC) {
+		if (direction != (*state)->direction) {
+			REASON_SET(reason, PFRES_SYNPROXY);
+			return (PF_SYNPROXY_DROP);
+		}
+		if (th->th_flags & TH_SYN) {
+			if (ntohl(th->th_seq) != (*state)->src.seqlo) {
+				REASON_SET(reason, PFRES_SYNPROXY);
+				return (PF_DROP);
+			}
+#ifdef __FreeBSD__
+			pf_send_tcp(NULL, (*state)->rule.ptr, pd->af, pd->dst,
+#else
+			pf_send_tcp((*state)->rule.ptr, pd->af, pd->dst,
+#endif
+			    pd->src, th->th_dport, th->th_sport,
+			    (*state)->src.seqhi, ntohl(th->th_seq) + 1,
+			    TH_SYN|TH_ACK, 0, (*state)->src.mss, 0, 1,
+			    0, NULL, NULL);
+			REASON_SET(reason, PFRES_SYNPROXY);
+			return (PF_SYNPROXY_DROP);
+		} else if (!(th->th_flags & TH_ACK) ||
+		    (ntohl(th->th_ack) != (*state)->src.seqhi + 1) ||
+		    (ntohl(th->th_seq) != (*state)->src.seqlo + 1)) {
+			REASON_SET(reason, PFRES_SYNPROXY);
+			return (PF_DROP);
+		} else if ((*state)->src_node != NULL &&
+		    pf_src_connlimit(state)) {
+			REASON_SET(reason, PFRES_SRCLIMIT);
+			return (PF_DROP);
+		} else
+			(*state)->src.state = PF_TCPS_PROXY_DST;
+	}
+	if ((*state)->src.state == PF_TCPS_PROXY_DST) {
+		struct pf_state_host *src, *dst;
+
+		if (direction == PF_OUT) {
+			src = &(*state)->gwy;
+			dst = &(*state)->ext;
+		} else {
+			src = &(*state)->ext;
+			dst = &(*state)->lan;
+		}
+		if (direction == (*state)->direction) {
+			if (((th->th_flags & (TH_SYN|TH_ACK)) != TH_ACK) ||
+			    (ntohl(th->th_ack) != (*state)->src.seqhi + 1) ||
+			    (ntohl(th->th_seq) != (*state)->src.seqlo + 1)) {
+				REASON_SET(reason, PFRES_SYNPROXY);
+				return (PF_DROP);
+			}
+			(*state)->src.max_win = MAX(ntohs(th->th_win), 1);
+			if ((*state)->dst.seqhi == 1)
+				(*state)->dst.seqhi = htonl(arc4random());
+#ifdef __FreeBSD__
+			pf_send_tcp(NULL, (*state)->rule.ptr, pd->af,
+			    &src->addr,
+#else
+			pf_send_tcp((*state)->rule.ptr, pd->af, &src->addr,
+#endif
+			    &dst->addr, src->port, dst->port,
+			    (*state)->dst.seqhi, 0, TH_SYN, 0,
+			    (*state)->src.mss, 0, 0, (*state)->tag, NULL, NULL);
+			REASON_SET(reason, PFRES_SYNPROXY);
+			return (PF_SYNPROXY_DROP);
+		} else if (((th->th_flags & (TH_SYN|TH_ACK)) !=
+		    (TH_SYN|TH_ACK)) ||
+		    (ntohl(th->th_ack) != (*state)->dst.seqhi + 1)) {
+			REASON_SET(reason, PFRES_SYNPROXY);
+			return (PF_DROP);
+		} else {
+			(*state)->dst.max_win = MAX(ntohs(th->th_win), 1);
+			(*state)->dst.seqlo = ntohl(th->th_seq);
+#ifdef __FreeBSD__
+			pf_send_tcp(NULL, (*state)->rule.ptr, pd->af, pd->dst,
+#else
+			pf_send_tcp((*state)->rule.ptr, pd->af, pd->dst,
+#endif
+			    pd->src, th->th_dport, th->th_sport,
+			    ntohl(th->th_ack), ntohl(th->th_seq) + 1,
+			    TH_ACK, (*state)->src.max_win, 0, 0, 0,
+			    (*state)->tag, NULL, NULL);
+#ifdef __FreeBSD__
+			pf_send_tcp(NULL, (*state)->rule.ptr, pd->af,
+			    &src->addr,
+#else
+			pf_send_tcp((*state)->rule.ptr, pd->af, &src->addr,
+#endif
+			    &dst->addr, src->port, dst->port,
+			    (*state)->src.seqhi + 1, (*state)->src.seqlo + 1,
+			    TH_ACK, (*state)->dst.max_win, 0, 0, 1,
+			    0, NULL, NULL);
+			(*state)->src.seqdiff = (*state)->dst.seqhi -
+			    (*state)->src.seqlo;
+			(*state)->dst.seqdiff = (*state)->src.seqhi -
+			    (*state)->dst.seqlo;
+			(*state)->src.seqhi = (*state)->src.seqlo +
+			    (*state)->dst.max_win;
+			(*state)->dst.seqhi = (*state)->dst.seqlo +
+			    (*state)->src.max_win;
+			(*state)->src.wscale = (*state)->dst.wscale = 0;
+			(*state)->src.state = (*state)->dst.state =
+			    TCPS_ESTABLISHED;
+			REASON_SET(reason, PFRES_SYNPROXY);
+			return (PF_SYNPROXY_DROP);
+		}
+	}
+
+	if (((th->th_flags & (TH_SYN|TH_ACK)) == TH_SYN) &&
+	    dst->state >= TCPS_FIN_WAIT_2 &&
+	    src->state >= TCPS_FIN_WAIT_2) {
+		if (pf_status.debug >= PF_DEBUG_MISC) {
+			printf("pf: state reuse ");
+			pf_print_state(*state);
+			pf_print_flags(th->th_flags);
+			printf("\n");
+		}
+		/* XXX make sure it's the same direction ?? */
+		(*state)->src.state = (*state)->dst.state = TCPS_CLOSED;
+		pf_unlink_state(*state);
+		*state = NULL;
+		return (PF_DROP);
+	}
+
+	if ((*state)->state_flags & PFSTATE_SLOPPY) {
+		if (pf_tcp_track_sloppy(src, dst, state, pd, reason) == PF_DROP)
+			return (PF_DROP);
+	} else {
+		if (pf_tcp_track_full(src, dst, state, kif, m, off, pd, reason,
+		    &copyback) == PF_DROP)
+			return (PF_DROP);
+	}
 
 	/* translate source/destination address, if necessary */
 	if (STATE_TRANSLATE(*state)) {
@@ -5533,8 +5650,9 @@ pf_test_state_icmp(struct pf_state **state, int direction, struct pfi_kif *kif,
 				copyback = 1;
 			}
 
-			if (!SEQ_GEQ(src->seqhi, seq) ||
-			    !SEQ_GEQ(seq, src->seqlo - (dst->max_win << dws))) {
+			if (!((*state)->state_flags & PFSTATE_SLOPPY) &&
+			    (!SEQ_GEQ(src->seqhi, seq) ||
+			    !SEQ_GEQ(seq, src->seqlo - (dst->max_win << dws)))) {
 				if (pf_status.debug >= PF_DEBUG_MISC) {
 					printf("pf: BAD ICMP %d:%d ",
 					    icmptype, pd->hdr.icmp->icmp_code);
@@ -7052,7 +7170,7 @@ pf_test(int dir, struct ifnet *ifp, struct mbuf **m0,
 
 done:
 	if (action == PF_PASS && h->ip_hl > 5 &&
-	    !((s && s->allow_opts) || r->allow_opts)) {
+	    !((s && s->state_flags & PFSTATE_ALLOWOPTS) || r->allow_opts)) {
 		action = PF_DROP;
 		REASON_SET(&reason, PFRES_IPOPTIONS);
 		log = 1;
@@ -7513,7 +7631,7 @@ pf_test6(int dir, struct ifnet *ifp, struct mbuf **m0,
 done:
 	/* handle dangerous IPv6 extension headers. */
 	if (action == PF_PASS && rh_cnt &&
-	    !((s && s->allow_opts) || r->allow_opts)) {
+	    !((s && s->state_flags & PFSTATE_ALLOWOPTS) || r->allow_opts)) {
 		action = PF_DROP;
 		REASON_SET(&reason, PFRES_IPOPTIONS);
 		log = 1;
diff --git a/sys/contrib/pf/net/pfvar.h b/sys/contrib/pf/net/pfvar.h
index 32e721c5733f..de175b1a0213 100644
--- a/sys/contrib/pf/net/pfvar.h
+++ b/sys/contrib/pf/net/pfvar.h
@@ -700,6 +700,7 @@ struct pf_rule {
 
 /* rule flags again */
 #define PFRULE_IFBOUND		0x00010000	/* if-bound */
+#define PFRULE_STATESLOPPY	0x00020000	/* sloppy state tracking */
 
 #define PFSTATE_HIWAT		10000	/* default state table size */
 #define PFSTATE_ADAPT_START	6000	/* default adaptive timeout start */
@@ -800,7 +801,9 @@ struct pf_state {
 	u_int8_t	 pad;
 #endif
 	u_int8_t	 log;
-	u_int8_t	 allow_opts;
+	u_int8_t	 state_flags;
+#define	PFSTATE_ALLOWOPTS	0x01
+#define	PFSTATE_SLOPPY		0x02
 	u_int8_t	 timeout;
 	u_int8_t	 sync_flags;
 #define	PFSTATE_NOSYNC	 0x01