Re: PCI: Revert "PCI: Add runtime PM support for PCIe ports"

[Peter Wu <peter@lekensteyn.nl>]

On Tue, Jan 03, 2017 at 05:11:23PM +0100, Lukas Wunner wrote:
> [cc += Dave Airlie:
> 
> Dave, we're about to lose support for newer Optimus laptops which use
> _PR3 to cut power to the discrete GPU because Bjorn Helgaas has queued
> a commit on his for-linus branch to remove runtime PM for PCIe ports.
> This fixes a regression on Kilian Singer's laptop on which locking the
> screen breaks USB and PS/2 input devices:  Mouse movements are still
> visible, but button or key presses no longer have any effect.  The GPU
> is powered down upon locking the screen and the current theory is that
> this causes the issues.]

(+cc Alex: this might affect amdgpu/radeon too.]

Bjorn, please reconsider the rpm patch. Reverting support would
introduce other regressions (see issues below) and make future
Thunderbolt work harder (according to Lukas). If Kilian's laptop has
issues, what about a "temporary" quirk?

> On Tue, Jan 03, 2017 at 04:15:47PM +0100, Peter Wu wrote:
> > The acpidump from the bug does not show known issues, it *looks* fine.
> > There have been other issues related to resuming power on newer Nvidia
> > hardware (https://bugs.freedesktop.org/show_bug.cgi?id=94725,
> > https://bugzilla.kernel.org/show_bug.cgi?id=156341) but there is not
> > much progress here.  (The last time I traced the PCIe register accesses
> > (via kprobes) and tried to disable some of those, it still did not help
> > with preventing the power issue.)
> 
> It seems that the _DSM method works on Kilian's laptop.  Would it be
> viable to default to _DSM if it's available, and only use _PR3 if not?

DSM should not be preferred when PR3 is available:

 - After MS introduced D3cold (PR3) support to Win8+, vendors are
   unlikely to test legacy DSM and the likelihood of breakage increases.
 - On one Lenovo laptop, the DSM method causes memory corruption while
   PR3 fixes this problem.
 - On some laptops, DSM keeps the fan on while PR3 stopped the noise.
 - On some laptops, DSM does not really power off the GPU and results in
   increased power consumption during runtime/system sleep. PR3 fully
   removes the power, as desired.
-- 
Kind regards,
Peter Wu
https://lekensteyn.nl