Re: [PATCH 3/3] Revert "ACPI,PCI,IRQ: remove SCI penalize function"

From: Thomas Gleixner
Date: Tue Oct 04 2016 - 10:27:04 EST


On Tue, 4 Oct 2016, Sinan Kaya wrote:

> On 10/4/2016 3:23 AM, Thomas Gleixner wrote:
> > On Sat, 1 Oct 2016, Sinan Kaya wrote:
> >
> >> This reverts commit 9e5ed6d1fb87 ("ACPI,PCI,IRQ: remove SCI penalize
> >> function"). SCI penalty API was replaced by the runtime penalty calculation
> >> based on the value of acpi_gbl_FADT.sci_interrupt.
> >
> > This does more than only reverting said commit ....
>
> The SCI function was removed in two steps (first refactor and then remove).
> I was trying to do the revert at one step. I can divide into two if it makes
> it better

No one step is fine. But this wants to be documented in the changelog.

> >> acpi_gbl_FADT.sci_interrupt type does not get updated at the right time
> >> for some platforms and results in incorrect penalty assignment for PCI
> >> IRQs as irq_get_trigger_type returns the wrong type.
> >
> > And the obvious question is: Why does irq_get_trigger_type() return the
> > wrong type?
>
> Here is some history:
>
> I now remember that Bjorn indicated the race condition possibility in this thread
> here.
>
> https://lkml.org/lkml/2016/3/8/640

> My understanding is that register_gsi function delivers the IRQ found in
> the ACPI table to the interrupt controller driver. Penalties are
> calculated before a link object is enabled to find out which interrupt
> has the least number of users. By the time penalties are calculated, the
> IRQ is not registered yet and it returns the wrong type.

Ok.

> >
> > What's the root cause of this problem? Your changelog does not tell
> > anything.
>
> If you are OK with the above description, I can add this to the commit message.

Yes please.

Thanks,

tglx