[PATCH 0/2] Battlemage D3Cold issues

[Rodrigo Vivi]

Let's disable D3cold by default on Battlemage for our production
trees while we investigate the current issues. More details on
the commit message of patch 1.

But also, let's not entirely disable d3cold on our development
trees. So, the second patch aims to temporarily live in our
topic branch while we investigate the issues with firmware
teams.

Main justification for the topic branch:

1. The combination of the BMG devices and Hosts that we currently
have available in our CI and in most of our development platforms
have a very good and stable behavior of D3cold -> D0 transitions.

2. If we disable everywhere in development environments and CI
we are sentenced to regressions that will be harder later to deal
with.

3. D3cold on DG2 worked very well on multiple combinations. Even
on Host boards where now BMG is failing, so there's still hope
that we will get them fixed.

4. This approach avois polution in our Kconfig options or in
module parameter.

5. Our developers facing the bad bugs can either disable at
runtime by
echo 0 > /sys/bus/pci/devices/<addr>/vram_d3cold_threshold
or disable at boot with linux parameter pcie_port_pm=off.

Another possible approach would involve a kind of quirk with
allow list of the hosts based on the host bridge pci id. But
that is hard to maintain.

Thanks,
Rodrigo.

Rodrigo Vivi (2):
  drm/xe/pm: Temporarily disable D3Cold on BMG
  drm/xe/pm: Re-enable D3Cold by default on BMG


-- 
2.48.1