Re: [RFC] conntrack event framework speedup

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



On Tue, Mar 15, 2022 at 10:41:21PM +0100, Florian Westphal wrote:
> Pablo Neira Ayuso <pablo@xxxxxxxxxxxxx> wrote:
> > > add new net.netfilter.nf_conntrack_events default mode: 2, autodetect.
> > 
> > Probably the sysctl entry does not make any sense anymore if you can
> > autodetect when there is a listener?
> 
> Hmmm, did not consider that.  I *think* we still want to allow to
> disable the feature because of xt_CT/nft_ct.
> 
> Someone might have nf_conntrack_events=0 and tehy could be using
> explicit configuration via templates (and then expect that only
> those flows that matched a '-j CT' rule generate events.

Maybe could you bump the ctnetlink_listeners counter when -j CT is
used with event filtering?

> > > in nfnetlink bind, inc pernet counter when event group is bound.
> > > in nfnetlink unbind, dec pernet counter when event group is unbound.
> > 
> > So you keep one counter per netlink group in netns area?
> 
> Rough sketch (doesn't compile/apply):
> 
> +++ b/net/netfilter/nfnetlink.c
> @@ -45,6 +45,7 @@ MODULE_DESCRIPTION("Netfilter messages via netlink socket");
>  static unsigned int nfnetlink_pernet_id __read_mostly;
>   
>   struct nfnl_net {
>   +  unsigned int ctnetlink_listeners;
>      struct sock *nfnl;
> 
>  static int nfnetlink_bind(struct net *net, int group)
>  {
>         const struct nfnetlink_subsystem *ss;
> @@ -691,11 +691,47 @@ static int nfnetlink_bind(struct net *net, int group)
>         if (!ss)
>                 request_module_nowait("nfnetlink-subsys-%d", type);
> +
> +       if (type == NFNL_SUBSYS_CTNETLINK) {
> +               struct nfnl_net *nfnlnet = nfnl_pernet(net);
> +
> +               nfnl_lock(NFNL_SUBSYS_CTNETLINK);
> +               nfnlnet->ctnetlink_listeners++;
> +               if (nfnlnet->ctnetlink_listeners == 1)
> +                       net->ct.ctnetlink_has_listener = true;
> +               nfnl_unlock(NFNL_SUBSYS_CTNETLINK);
> 
> and then check 'net->ct.ctnetlink_has_listener' when allocating
> a new conntrack.

LGTM.

> > > a prototype.
> > 
> > There is also setsockopt() to subscribe to netlink groups, you might
> > need to extend netlink_kernel_cfg to deal with this case too?
> 
> afaics the netlink_bind callback is invoked for subscriptions via setsockopt too,
> so that angle shouw be covered.

good.



[Index of Archives]     [Netfitler Users]     [Berkeley Packet Filter]     [LARTC]     [Bugtraq]     [Yosemite Forum]

  Powered by Linux