Lack of _HPX after reset, DPC

Lack of _HPX after reset, DPC

[Bjorn Helgaas]

I think Linux is missing some important _HPX functionality.  Per ACPI
r6.5, sec 6.2.9,

  OSPM uses the information returned by _HPX to determine how to
  configure PCI Functions that are hot-plugged into the system, to
  configure Functions not configured by the platform firmware during
  initial system boot, and to configure Functions any time they lose
  configuration space settings (e.g. OSPM issues a Secondary Bus
  Reset/Function Level Reset or Downstream Port Containment is
  triggered).

Linux currently *does* process _HPX for hot-added devices.

The spec doesn't call it out for boot-time enumeration, except for
"Functions not configured by the platform firmware during initial
system boot", but Linux does process _HPX for all devices enumerated
at boot-time because it's not clear how to identify devices that
weren't configured by firmware.

But AFAICT, Linux does not do anything with _HPX in the device reset,
AER, or DPC flows.  I don't have any problem reports that I can say
are caused by lack of _HPX after reset, but it seems like something we
should fix.

If we could find a system with _HPX that does something interesting,
we might be able to demonstrate a defect by looking at a device before
and after a reset.

Bjorn

Re: Lack of _HPX after reset, DPC

[Kai-Heng Feng]

Hi Bjorn,

On Tue, Jul 16, 2024 at 5:45 AM Bjorn Helgaas <helgaas@kernel.org> wrote:
>
> I think Linux is missing some important _HPX functionality.  Per ACPI
> r6.5, sec 6.2.9,
>
>   OSPM uses the information returned by _HPX to determine how to
>   configure PCI Functions that are hot-plugged into the system, to
>   configure Functions not configured by the platform firmware during
>   initial system boot, and to configure Functions any time they lose
>   configuration space settings (e.g. OSPM issues a Secondary Bus
>   Reset/Function Level Reset or Downstream Port Containment is
>   triggered).
>
> Linux currently *does* process _HPX for hot-added devices.
>
> The spec doesn't call it out for boot-time enumeration, except for
> "Functions not configured by the platform firmware during initial
> system boot", but Linux does process _HPX for all devices enumerated
> at boot-time because it's not clear how to identify devices that
> weren't configured by firmware.
>
> But AFAICT, Linux does not do anything with _HPX in the device reset,
> AER, or DPC flows.  I don't have any problem reports that I can say
> are caused by lack of _HPX after reset, but it seems like something we
> should fix.

The consumer grade devices I checked don't have _HPX in the ACPI ASL.
So I can't really delve into it any further.

Kai-Heng

>
> If we could find a system with _HPX that does something interesting,
> we might be able to demonstrate a defect by looking at a device before
> and after a reset.
>
> Bjorn