Re: [PATCH nf 1/1] netfilter: bridge: release template ct on non-IP path

Jakub Kicinski <[email protected]> Fri, 31 Jul 2026 16:41:09 -0700
Newsgroups dev.linux.lists.bridge,org.kernel.vger.netdev,org.kernel.vger.netfilter-devel
Message-ID <[email protected]>
On Fri, 31 Jul 2026 17:47:40 +0200 Pablo Neira Ayuso wrote:
> On Fri, Jul 31, 2026 at 08:02:33AM -0700, Jakub Kicinski wrote:
> > On Fri, 31 Jul 2026 14:36:53 +0800 Zhiling Zou wrote:  
> > > A bridge nftables ct zone set rule can attach a conntrack template to
> > > an skb before nf_ct_bridge_pre() sees it. For non-IPv4 and non-IPv6
> > > EtherTypes, nf_ct_bridge_pre() currently overwrites skb->_nfct with
> > > IP_CT_UNTRACKED without releasing the existing template reference.  
> > 
> > could be a coincidence but around the time this landed in our test tree
> > the nft_queue.sh test became more flaky.  
> 
> Maybe this patch could be causing it?
> 
> [nf,v3,1/1] netfilter: validate L4 headers after userspace packet writes
> https://patchwork.ozlabs.org/project/netfilter-devel/patch/587830a57a593c4b621f7e8e94b2ddc73e6becb0.1785468829.git.zhilinz@nebusec.ai/
> 
> This is related to nfqueue specifically and it also was posted a few
> hours before the same day?

Yes, that makes more sense. There were some passes after that one go
applied but it's a flake so probably just luck