Re: [PATCH nf-next 1/2] netfilter: kill the fake untracked conntrack objects

[Pablo Neira Ayuso <pablo@netfilter.org>]

On Sat, Apr 15, 2017 at 11:43:15AM +0200, Florian Westphal wrote:
> Pablo Neira Ayuso <pablo@netfilter.org> wrote:
> > Hi Florian,
> > 
> > On Fri, Apr 14, 2017 at 08:31:08PM +0200, Florian Westphal wrote:
> > [...]
> > > resurrect an old patch from Pablo Neira to remove the untracked objects.
> > 
> > Thanks for doing so :)
> > 
> > > diff --git a/include/uapi/linux/netfilter/nf_conntrack_common.h b/include/uapi/linux/netfilter/nf_conntrack_common.h
> > > index 6a8e33dd4ecb..b4a0a1940118 100644
> > > --- a/include/uapi/linux/netfilter/nf_conntrack_common.h
> > > +++ b/include/uapi/linux/netfilter/nf_conntrack_common.h
> > > @@ -28,12 +28,14 @@ enum ip_conntrack_info {
> > >  	/* only for userspace compatibility */
> > >  #ifndef __KERNEL__
> > >  	IP_CT_NEW_REPLY = IP_CT_NUMBER,
> > > +#else
> > > +	IP_CT_UNTRACKED = 7,
> > 
> > This seems to be exposed via nfnetlink_queue conntrack support.
> 
> Yet another argument for removing this, its a bug.
> (missing !nf_ct_is_untracked() check).
> 
> Probably should also check !nf_ct_is_template() there.

We already do there. So no problem :)

> > > @@ -94,7 +96,7 @@ enum ip_conntrack_status {
> > >  	IPS_TEMPLATE_BIT = 11,
> > >  	IPS_TEMPLATE = (1 << IPS_TEMPLATE_BIT),
> > >  
> > > -	/* Conntrack is a fake untracked entry */
> > > +	/* Conntrack is a fake untracked entry.  Obsolete and not used anymore */
> > >  	IPS_UNTRACKED_BIT = 12,
> > >  	IPS_UNTRACKED = (1 << IPS_UNTRACKED_BIT),
> > 
> > I wonder if we should just set this via ctnetlink, so the interface
> > keeps stable. Even if we don't use this bit anymore. These are also
> > exposed via ctnetlink CTA_STATUS.
> 
> Not following, sorry (still jetlagged).
> ctnetlink_new_conntrack() always places the ct into the hashes, so its
> not 'untracked'.
> 
> Setting and/or changing IPS_UNTRACKED_BIT via ctnetlink should be
> disallowed.
> 
> Is that what you meant?

Just got confused in this nfnetlink_queue path, this looks good! :)