netdev - [PATCH net-next 09/14] netfilter: nft_flow_offload: clear tcp MAXACK flag before moving to slowpath

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite for Android: free password hash cracker in your pocket

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-Id: <20250116171902.1783620-10-pablo@netfilter.org>
Date: Thu, 16 Jan 2025 18:18:57 +0100
From: Pablo Neira Ayuso <pablo@...filter.org>
To: netfilter-devel@...r.kernel.org
Cc: davem@...emloft.net,
	netdev@...r.kernel.org,
	kuba@...nel.org,
	pabeni@...hat.com,
	edumazet@...gle.com,
	fw@...len.de
Subject: [PATCH net-next 09/14] netfilter: nft_flow_offload: clear tcp MAXACK flag before moving to slowpath

From: Florian Westphal <fw@...len.de>

This state reset is racy, no locks are held here.

Since commit
8437a6209f76 ("netfilter: nft_flow_offload: set liberal tracking mode for tcp"),
the window checks are disabled for normal data packets, but MAXACK flag
is checked when validating TCP resets.

Clear the flag so tcp reset validation checks are ignored.

Signed-off-by: Florian Westphal <fw@...len.de>
Signed-off-by: Pablo Neira Ayuso <pablo@...filter.org>
---
 net/netfilter/nf_flow_table_core.c | 14 ++++++++++++--
 1 file changed, 12 insertions(+), 2 deletions(-)

diff --git a/net/netfilter/nf_flow_table_core.c b/net/netfilter/nf_flow_table_core.c
index df72b0376970..bdde469bbbd1 100644
--- a/net/netfilter/nf_flow_table_core.c
+++ b/net/netfilter/nf_flow_table_core.c
@@ -161,10 +161,20 @@ void flow_offload_route_init(struct flow_offload *flow,
 }
 EXPORT_SYMBOL_GPL(flow_offload_route_init);
 
-static void flow_offload_fixup_tcp(struct ip_ct_tcp *tcp)
+static void flow_offload_fixup_tcp(struct nf_conn *ct)
 {
+	struct ip_ct_tcp *tcp = &ct->proto.tcp;
+
+	spin_lock_bh(&ct->lock);
+	/* Conntrack state is outdated due to offload bypass.
+	 * Clear IP_CT_TCP_FLAG_MAXACK_SET, otherwise conntracks
+	 * TCP reset validation will fail.
+	 */
 	tcp->seen[0].td_maxwin = 0;
+	tcp->seen[0].flags &= ~IP_CT_TCP_FLAG_MAXACK_SET;
 	tcp->seen[1].td_maxwin = 0;
+	tcp->seen[1].flags &= ~IP_CT_TCP_FLAG_MAXACK_SET;
+	spin_unlock_bh(&ct->lock);
 }
 
 static void flow_offload_fixup_ct(struct nf_conn *ct)
@@ -176,7 +186,7 @@ static void flow_offload_fixup_ct(struct nf_conn *ct)
 	if (l4num == IPPROTO_TCP) {
 		struct nf_tcp_net *tn = nf_tcp_pernet(net);
 
-		flow_offload_fixup_tcp(&ct->proto.tcp);
+		flow_offload_fixup_tcp(ct);
 
 		timeout = tn->timeouts[ct->proto.tcp.state];
 		timeout -= tn->offload_timeout;
-- 
2.30.2