[PATCH 20/38] netns ct: NOTRACK in netns

Alexey Dobriyan adobriyan at gmail.com
Sun Aug 24 03:43:13 PDT 2008


On Sat, Aug 23, 2008 at 08:35:07PM -0400, Jan Engelhardt wrote:
> 
> On Friday 2008-08-22 07:30, adobriyan at gmail.com wrote:
> >
> >We wait for untracked ct refcount to drop to 1 back:
> >
> >	/* wait until all references to nf_conntrack_untracked are dropped */
> >	while (atomic_read(&nf_conntrack_untracked.ct_general.use) > 1)
> >		schedule();
> >
> >Consequently it should be one per netns, otherwise netns A can prevent
> >netns B from stopping.
> >
> 
> But nf_conntrack_cleanup is not per netns, is it?

That's because nf_conntrack_cleanup() is _code_.

If netns A actively uses NOTRACK, untracked ct refcount will be bumped.
And netns B which haven't used NOTRACK at all will wait for netns A to
stop using NOTRACK potentially indefinitely.

> At least I do not think it should be.



More information about the Containers mailing list