The root cause is an ICMPv6 packet hits an untracked connection and inherits a mark from the previous packet to which it is not related, IMO it doesn't make sense to inherit mark for nf_conntrack_untracked.
Signed-off-by: XU Tianwen <evan.xu.tian...@gmail.com> --- net/ipv6/netfilter/nf_conntrack_proto_icmpv6.c | 3 +++ 1 file changed, 3 insertions(+) diff --git a/net/ipv6/netfilter/nf_conntrack_proto_icmpv6.c b/net/ipv6/netfilter/nf_conntrack_proto_icmpv6.c index f5a61bc..1be9000 100644 --- a/net/ipv6/netfilter/nf_conntrack_proto_icmpv6.c +++ b/net/ipv6/netfilter/nf_conntrack_proto_icmpv6.c @@ -223,6 +223,9 @@ icmpv6_error(struct net *net, struct nf_conn *tmpl, if (type >= 0 && type < sizeof(noct_valid_new) && noct_valid_new[type]) { skb->nfct = &nf_ct_untracked_get()->ct_general; +#ifdef CONFIG_NF_CONNTRACK_MARK + ((struct nf_conn *)skb->nfct)->mark = 0; +#endif skb->nfctinfo = IP_CT_NEW; nf_conntrack_get(skb->nfct); return NF_ACCEPT; -- 2.1.2