Re: [PATCH v4 1/2] PCI: Add tango MSI controller support

From: Bjorn Helgaas
Date: Wed May 31 2017 - 16:04:38 EST


On Wed, May 31, 2017 at 09:27:50PM +0200, Mason wrote:
> On 31/05/2017 21:12, Bjorn Helgaas wrote:
> > On Wed, May 31, 2017 at 02:00:37PM -0500, Bjorn Helgaas wrote:
> >> On Wed, May 31, 2017 at 08:49:04PM +0200, Mason wrote:
> >>> On 31/05/2017 19:34, Bjorn Helgaas wrote:
> >>> ...
> >>
> >>>> This would be more an IRQ patch than a PCI patch, but if I were
> >>>> reviewing it, I would look for assurance that *all* the no-op
> >>>> .irq_set_affinity callbacks were cleaned up, not just those in
> >>>> drivers/pci/host.
> >>>
> >>> Are you saying the patch is *wrong* if not all "do-nothing"
> >>> callbacks are cleaned up?
> >>
> >> I'm saying that (1) this probably wouldn't be applied via the PCI
> >> tree, and (2) if it *were* applied via PCI, I would ask that all the
> >> no-op callbacks were cleaned up at the same time.
> >>
> >> Huh, that sounds a lot like what I wrote above. Was I unclear?
> >
> > I'm afraid this sounded snarky, which isn't my intention. It seems
> > like there's a useful patch here, and I didn't want to see it get
> > ignored for lack of following the usual process. If this is all
> > obvious to you, my apologies and please ignore my suggestion.
>
> Thanks for clearing things up. I had indeed assumed from
> your first reply that the patch was pointless.
>
> Writing a script locating all candidates will be an
> interesting exercise.

Cscope only sees 94 definitions of irq_set_affinity. I know *I* could
never write a script faster than looking at them manually. While
doing that, I noticed irq_chip_set_affinity_parent(), which is used in
14 cases and appears similar to your patch.