Re: CPU Hotplug rework

["Paul E. McKenney" <paulmck@linux.vnet.ibm.com>]

On Tue, Apr 10, 2012 at 08:09:59PM -0400, Steven Rostedt wrote:
> On Sat, 2012-04-07 at 01:22 +0530, Srivatsa S. Bhat wrote:
> 
> > Wrapping only do_setup() within get/put_online_cpus() wouldn't serve our
> > purpose, since the race with CPU Hotplug would still exist, just like
> > before. So, let's consider what happens when we wrap both the functions
> > within get/put_online_cpus():
> > 
> > 	get_online_cpus();
> > 	register_cpu_notifier(nb, mask);
> > 	do_setup(mask);
> > 	put_online_cpus();
> > 
> > Unfortunately this leads to an ABBA deadlock (see below).
> >  
> 
> Just to throw out the stupid silly approach.
> 
> What about creating a "__register_cpu_notifier()" that just does:
> 
> int __ref __register_cpu_notifier(struct notifier_block *nb)
> {
> 	return raw_notifier_chain_register(&cpu_chain, nb);
> }
> 
> Also making cpu_maps_update_begin/done() global (and probably rename
> them).
> 
> and then in the above code do:
> 
> 	cpu_maps_update_begin();
> 	__register_cpu_notifier(nb);
> 	do_setup();
> 	cpu_maps_update_done();
> 
> 
> Just saying,

That does have some attractive properties, now that you mention it.  ;-)

							Thanx, Paul