Re: [RFC] Suspicious bug in module refcounting

[Rusty Russell <rusty@rustcorp.com.au>]

On Tuesday 10 February 2009 21:01:37 Michal Hocko wrote:
> On Tue 10-02-09 13:45:07, Rusty Russell wrote:
> > On Tuesday 10 February 2009 01:48:31 Michal Hocko wrote:
> > > Based on this change, would it make sense to update sys_accept to change
> > > __module_get to try_module_get like in the following patch?
> > 
> > I don't think so:
> > 
> > >  	/*
> > > -	 * We don't need try_module_get here, as the listening socket (sock)
> > > -	 * has the protocol module (sock->ops->owner) held.
> > > +	 * Socket's owner cannot be in unloading path because there
> > > +	 * must be at least one listening reference
> > >  	 */
> > > -	__module_get(newsock->ops->owner);
> > > +	if (unlikely(!try_module_get(newsock->ops->owner)))
> > > +		BUG();
> > 
> > rmmod --wait can make try_module_get fail even if the reference count isn't
> > zero.  
> 
> OK, I though that rmmod --wait waits for refcount==0 and then changes
> the state.

No, it has to stop all future use, otherwise it's useless under load.

Cheers,
Rusty.