Skip to contentSkip to conversation

netfilter: conntrack: handle tcp challenge acks during connection reuse

PATCH New Florian Westphal 1 participant 1 message Jan 11, 2023

Subsystems (2)
lore
Share & tools
lore

Florian Westphal

author maintainer
#0 lore raw

PATCH 1/1 netfilter: conntrack: handle tcp challenge acks during connection reuse

When a connection is re-used, following can happen:
[ connection starts to close, fin sent in either direction ]

Quoted 1 line

syn # initator quickly reuses connection

 < ack   # peer sends a challenge ack
 > rst   # rst, sequence number == ack_seq of previous challenge ack
 > syn   # this syn is expected to pass

Problem is that the rst will fail window validation, so it gets tagged as invalid.

If ruleset drops such packets, we get repeated syn-retransmits until initator gives up or peer starts responding with syn/ack.

Before the commit indicated in the "Fixes" tag below this used to work:

The challenge-ack made conntrack re-init state based on the challenge ack itself, so the following rst would pass window validation.

Add challenge-ack support: If we get ack for syn, record the ack_seq, and then check if the rst sequence number matches the last ack number seen in reverse direction.

Fixes c7aab4f17021 ("netfilter: nf_conntrack_tcp: re-init for syn packets only")
---
 net/netfilter/nf_conntrack_proto_tcp.c | 15 +++++++++++++++
 1 file changed, 15 insertions(+)
net/netfilter/nf_conntrack_proto_tcp.c +15 -0
index 656631083177..3ac1af6f59fc 100644
@@ -1068,6 +1068,13 @@ int nf_conntrack_tcp_packet(struct nf_conn *ct,
 				ct->proto.tcp.last_flags |=
 					IP_CT_EXP_CHALLENGE_ACK;
 		}
+
+		/* possible challenge ack reply to syn */
+		if (old_state == TCP_CONNTRACK_SYN_SENT &&
+		    index == TCP_ACK_SET &&
+		    dir == IP_CT_DIR_REPLY)
+			ct->proto.tcp.last_ack = ntohl(th->ack_seq);
+
 		spin_unlock_bh(&ct->lock);
 		nf_ct_l4proto_log_invalid(skb, ct, state,
 					  "packet (index %d) in dir %d ignored, state %s",
@@ -1193,6 +1200,14 @@ int nf_conntrack_tcp_packet(struct nf_conn *ct,
 			 * segments we ignored. */
 			goto in_window;
 		}
+
+		/* Reset in response to a challenge-ack we let through earlier */
+		if (old_state == TCP_CONNTRACK_SYN_SENT &&
+		    ct->proto.tcp.last_index == TCP_ACK_SET &&
+		    ct->proto.tcp.last_dir == IP_CT_DIR_REPLY &&
+		    ntohl(th->seq) == ct->proto.tcp.last_ack)
+			goto in_window;
+
 		break;
 	default:
 		/* Keep compilers happy. */
Signature · 1 line
2.38.2