Re: [PATCH v2 17/17] irq: remove handle_domain_{irq,nmi}()

[Lukas Wunner <lukas@xxxxxxxxx>]

On Wed, May 11, 2022 at 09:23:56AM +0100, Mark Rutland wrote:
> On Wed, May 11, 2022 at 12:52:29AM +0200, Thomas Gleixner wrote:
> > +	/* USB interrupts are received in softirq (tasklet) context.
> > +	 * Switch to hardirq context to make genirq code happy.
> > +	 */
> > +	local_irq_save(flags);
> > +	__irq_enter_raw();
> > +
> >  	if (intdata & INT_ENP_PHY_INT_)
> > -		;
> > +		generic_handle_domain_irq(pdata->irqdomain, PHY_HWIRQ);
> 
> Agreed. IIUC everyone agrees the __irq_enter_raw() usage is a hack,
> but what's not clear is what we *should* do
> 
> I suspect that given we have generic_handle_irq_safe() for situations
> like this we should add a generic_handle_domain_irq_safe(), and use
> that in this driver?
> That way we can keep the `WARN_ON_ONCE(!in_hardirq())` in
> generic_handle_domain_irq().

Thomas applied 792ea6a074ae ("genirq: Remove WARN_ON_ONCE() in
generic_handle_domain_irq()") tonight:

http://git.kernel.org/tip/tip/c/792ea6a074ae

That allows me to drop the controversial __irq_enter_raw()
and thus unblock my smsc95xx series.

generic_handle_domain_irq_safe() would merely be a wrapper for
generic_handle_domain_irq() which disables local interrupts.
Then I wouldn't have to do that in smsc95xx.c.  IMHO that's a
cosmetic improvement, though I'll be happy to provide a patch
if desired?

Thanks,

Lukas