Re: [patch 1/5] genirq: Delay deactivation in free_irq()

[Thomas Gleixner]

Marc,

On Fri, 28 Jun 2019, Marc Zyngier wrote:
> > --- a/kernel/irq/autoprobe.c
> > +++ b/kernel/irq/autoprobe.c
> > @@ -1699,6 +1700,7 @@ static struct irqaction *__free_irq(stru
> >  	/* If this was the last handler, shut down the IRQ line: */
> >  	if (!desc->action) {
> >  		irq_settings_clr_disable_unlazy(desc);
> > +		/* Only shutdown. Deactivate after synchronize_irq() */
> 
> synchronize_hardirq()

Correct.

> >  
> >  	irq_release_resources(desc);
> The patch makes complete sense, so this comment is only a nit, feel free 
> to ignore me: I find it a bit odd that irq_shutdown() doesn't do the 
> full thing anymore, given that it is the "canonical" function, and that 
> __free_irq is the only one that has special needs. Suggestion below.

We have pretty much the same thing with startup, so I made it symmetric.

> Irrespective of which version you prefer:
> 
> Reviewed-by: Marc Zyngier <marc.zyngier@xxxxxxx>

Thanks for looking into this!

       tglx