Re: [PATCH] PCI: Move test of INTx masking to pci_setup_device

[Piotr Gregor <piotrgregor@rsyncme.org>]

On Sat, Jun 17, 2017 at 08:14:33PM -0500, Bjorn Helgaas wrote:
> On Sat, Jun 17, 2017 at 06:45:44PM +0100, Piotr Gregor wrote:
> > Hi Bjorn,
> > 
> > The pci_cfg_access_lock is most likely not needed there.
> > The assignment by return type is indeed preferred in this case.
> > 
> > However, you have changed the meaning of returned boolean information
> > by pci_intx_mask_broken leaving pci_intx_mask_supported unchanged.
> > The test should be: 
> > 
> >     if (new != toggle) /* the test failed */
> > 	        return 1;
> > 	return 0;
> 
> Oh, you're absolutely right, thanks for catching that!  I updated my
> pci/enumeration branch.
> 
> > Regarding v2.3 - do you think it is worth to apply the check
> > so we would have something like
> > 
> >     if ((new == toggle) || PCI_VERSION_PRIOR_TO_23) /* test OK or PCI prior to r2.3 */
> > 	        return 0;
> > 	return 1;
> 
> I'm not sure how to test for r2.3 compliance.  But even if we could, I
> guess I think the current code is probably better because it actually
> checks the property we care about, not a spec revision that is one
> step removed from the property.
> 
> Bjorn

Hi Bjorn,

You are right, having

        if ((new == toggle) || PCI_VERSION_PRIOR_TO_23) /* test OK or PCI prior to r2.3 */
		        return 0;
        return 1;

would be incorrect, as if new != toggle then PCI_COMMAND_INTX_DISABLE is not writable
so INTx masking support should be considered broken (regardless of PCI version).

Piotr