Re: [PATCH] netfilter: fix ->nfnl NULL oops

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

On Mon, Nov 21, 2011 at 11:50:36PM +0000, Alex Bligh wrote:
> 
> 
> --On 22 November 2011 00:39:42 +0100 Pablo Neira Ayuso
> <pablo@netfilter.org> wrote:
> 
> >I didn't try with this script, but the problem can be easily
> >triggered with:
> >
> >0) make sure nf_conntrack_netlink and nf_conntrack_ipv4 are loaded.
> >1) container is started.
> >2) connect to it via lxc-console.
> >3) generate some traffic with the container to create some conntrack
> >   entries in its table.
> >4) stop the container: hit the oops.
> >
> >I've been testing the patch that I proposed with this recipe, now it
> >works fine.
> >
> >I'll pass my patch for 3.2-rc soon, in case you want to make further
> >testing of it.
> 
> I will try to, when I have a minute. Occasionally I was getting a
> double oops before. My theory at the time was that this was to do
> with passing traffic /as/ the container was being destroyed. I have
> little to substantiate that, but that was the reason for the perl
> script (which really just runs a ping as the container is being
> destroyed, which ensures there are conntrack entries).

I think my fix is OK but more validation is always welcome, so I'd
appreciate if you validate this with your script ;-).