Re: [PATCH v1 1/1] mfd: intel-lpss: Switch over to MSI interrupts

[Andy Shevchenko]

On Tue, Mar 26, 2024 at 05:09:53PM -0500, Bjorn Helgaas wrote:
> On Tue, Mar 26, 2024 at 11:22:16PM +0200, Andy Shevchenko wrote:
> > On Tue, Mar 26, 2024 at 04:01:07PM -0500, Bjorn Helgaas wrote:
> > > On Tue, Mar 26, 2024 at 06:21:47PM +0200, Andy Shevchenko wrote:
> > > > On Mon, Mar 25, 2024 at 04:19:15PM -0500, Bjorn Helgaas wrote:
> > > > > On Tue, Mar 12, 2024 at 06:59:05PM +0200, Andy Shevchenko wrote:

..

> > > > > > -	ret = pci_alloc_irq_vectors(pdev, 1, 1, PCI_IRQ_LEGACY);
> > > > > > +	ret = pci_alloc_irq_vectors(pdev, 1, 1, PCI_IRQ_ALL_TYPES);
> > > > > >  	if (ret < 0)
> > > > > >  		return ret;
> > > > > 
> > > > > I guess at least some of these devices do support INTx, since we
> > > > > always used INTx previously, right?
> > > > > 
> > > > > There are a bunch of bug reports complaining about a lack of _PRT
> > > > > entries for them, e.g., these from
> > > > > https://bugzilla.kernel.org/show_bug.cgi?id=212261#c24:
> > > > 
> > > > But this is not related to my patch, and the mentioned bug report seems about
> > > > all AMD and Intel platforms.
> > > > 
> > > > Can you, please, elaborate what the relation to my patch?
> > > 
> > > Right, sorry I didn't make that clear; I didn't mean that it was
> > > related to your patch.  I was just looking at this old bug report
> > > about not being able to figure out INTx routing.
> > > 
> > > Your patch had to do with interrupts, so I just wondered whether you
> > > had insight into whether these devices actually used INTx.  My guess
> > > is that at least some of them *do* use INTx, because prior to your
> > > patch, the driver *only* tried to use INTx.
> > > 
> > > If it happend that they never use INTx, but advertise INTA via
> > > Interrupt Pin, I think that would be a device defect that we might
> > > consider a quirk for.
> > > 
> > > If they *do* use INTx, and the _PRT doesn't tell us how it's routed, I
> > > think that would be a firmware defect, and ... I dunno what we would
> > > do.  I guess just avoid using INTx because we don't know where the
> > > interupt goes.
> > 
> > Okay, so the revert after all is not required, do you agree?
> 
> Yes, I agree!  No indication of problems with your patch, AFAICS.
> 
> If you have any opinions or ideas on the "can't derive routing for PCI
> INT A" stuff, I'd still be interested, because it really annoys users.

Got it, thank you for clarification!

-- 
With Best Regards,
Andy Shevchenko