Re: [Intel-gfx] [PATCH 1/4] drm/xe: Only set PCI d3cold_allowed when we are really allowing.

[Rodrigo Vivi <rodrigo.vivi@intel.com>]

On Tue, Jul 25, 2023 at 01:08:11AM -0400, Gupta, Anshuman wrote:
> 
> 
> > -----Original Message-----
> > From: Vivi, Rodrigo <rodrigo.vivi@intel.com>
> > Sent: Saturday, July 22, 2023 12:30 AM
> > To: Gupta, Anshuman <anshuman.gupta@intel.com>
> > Cc: intel-gfx@lists.freedesktop.org
> > Subject: Re: [PATCH 1/4] drm/xe: Only set PCI d3cold_allowed when we are
> > really allowing.
> > 
> > On Fri, Jul 21, 2023 at 03:39:35AM -0400, Gupta, Anshuman wrote:
> > >
> > >
> > > > -----Original Message-----
> > > > From: Vivi, Rodrigo <rodrigo.vivi@intel.com>
> > > > Sent: Friday, July 21, 2023 2:34 AM
> > > > To: intel-gfx@lists.freedesktop.org
> > > > Cc: Vivi, Rodrigo <rodrigo.vivi@intel.com>; Gupta, Anshuman
> > > > <anshuman.gupta@intel.com>
> > > > Subject: [PATCH 1/4] drm/xe: Only set PCI d3cold_allowed when we are
> > > > really allowing.
> > > >
> > > > First of all it was strange to see:
> > > > if (allowed) {
> > > > ...
> > > > } else {
> > > >    D3COLD_ENABLE
> > > > }
> > > >
> > > > But besides this misalignment, let's also use the pci d3cold_allowed
> > > > useful to us and know that we are not really allowing d3cold.
> > > >
> > > > Cc: Anshuman Gupta <anshuman.gupta@intel.com>
> > > > Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
> > > > ---
> > > >  drivers/gpu/drm/xe/xe_pci.c | 3 +--
> > > >  1 file changed, 1 insertion(+), 2 deletions(-)
> > > >
> > > > diff --git a/drivers/gpu/drm/xe/xe_pci.c
> > > > b/drivers/gpu/drm/xe/xe_pci.c index 78df43c20cd2..0c4051f4f746
> > > > 100644
> > > > --- a/drivers/gpu/drm/xe/xe_pci.c
> > > > +++ b/drivers/gpu/drm/xe/xe_pci.c
> > > > @@ -794,6 +794,7 @@ static int xe_pci_runtime_suspend(struct device
> > > > *dev)
> > > >  	pci_save_state(pdev);
> > > >
> > > >  	if (xe->d3cold.allowed) {
> > > > +		d3cold_toggle(pdev, D3COLD_ENABLE);
> > > >  		pci_disable_device(pdev);
> > > >  		pci_ignore_hotplug(pdev);
> > > >  		pci_set_power_state(pdev, PCI_D3cold); @@ -823,8 +824,6
> > @@ static
> > > > int xe_pci_runtime_resume(struct device *dev)
> > > >  			return err;
> > > >
> > > >  		pci_set_master(pdev);
> > > > -	} else {
> > > > -		d3cold_toggle(pdev, D3COLD_ENABLE);
> > > >  	}
> > > During s2idle , d3cold may get disabled if won't restore it's state in runtime
> > resume.
> > 
> > I always forget about that case... please disregard this patch.
> I think , we can have this patch.
> I realized that system wide suspend/resume d3cold need some brainstorming.
> If device is already in runtime suspend state during system wide suspend, PM core may smartly skip device suspend/resume callback on Xe driver.
> This wasn't the case on I915 driver as it explicitly disables those smart optimization by 
> dev_pm_set_driver_flags(kdev, DPM_FLAG_NO_DIRECT_COMPLETE). 

so, it looks that we do need this as well anyway.

1. because we might not be in the runtime-d3cold, but on the system suspend
   we will lose power, hence the eviction/restore needs to happen.
2. because of the hda audio as documented in i915: "becaue the HDA driver
   may require us to enable the audio power domain during system suspend."

then, on device suspend we enable the d3cold and disable again on
device resume.

> Thanks,
> Anshuman.
> Thanks,
> Anshuman Gupta.
> > 
> > > Thanks,
> > > Anshuman Gupta.
> > > >
> > > >  	return xe_pm_runtime_resume(xe);
> > > > --
> > > > 2.41.0
> > >