[PATCH] drm/amd: Don't reset dGPUs if the system is going to s2idle

[Alex Deucher]

On Tue, May 17, 2022 at 10:06 AM Limonciello, Mario
<Mario.Limonciello at amd.com> wrote:
>
> [Public]
>
>
>
> > -----Original Message-----
> > From: Alex Deucher <alexdeucher at gmail.com>
> > Sent: Tuesday, May 17, 2022 08:43
> > To: Limonciello, Mario <Mario.Limonciello at amd.com>
> > Cc: amd-gfx list <amd-gfx at lists.freedesktop.org>
> > Subject: Re: [PATCH] drm/amd: Don't reset dGPUs if the system is going to
> > s2idle
> >
> > On Tue, May 17, 2022 at 9:39 AM Mario Limonciello
> > <mario.limonciello at amd.com> wrote:
> > >
> > > An A+A configuration on ASUS ROG Strix G513QY proves that the ASIC
> > > reset for handling aborted suspend can't work with s2idle.
> > >
> > > This functionality was introduced in commit daf8de0874ab5b
> > ("drm/amdgpu:
> > > always reset the asic in suspend (v2)").  A few other commits have
> > > gone on top of the ASIC reset, but this still doesn't work on the A+A
> > > configuration in s2idle.
> > >
> > > Avoid doing the reset on dGPUs specifically when using s2idle.
> > >
> > > Fixes: daf8de0874ab5b ("drm/amdgpu: always reset the asic in suspend
> > (v2)")
> > > Link:
> > https://nam11.safelinks.protection.outlook.com/?url=https%3A%2F%2Fgitla
> > b.freedesktop.org%2Fdrm%2Famd%2F-
> > %2Fissues%2F2008&data=05%7C01%7Cmario.limonciello%40amd.com%
> > 7C38a880b8d03147194bb608da380b3142%7C3dd8961fe4884e608e11a82d994
> > e183d%7C0%7C0%7C637883917968950850%7CUnknown%7CTWFpbGZsb3d8
> > eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3
> > D%7C3000%7C%7C%7C&sdata=lNeeucWFganu9GLey2YAqXfgbYT4DUBb
> > fQA6XuwGslA%3D&reserved=0
> > > Signed-off-by: Mario Limonciello <mario.limonciello at amd.com>
> >
> > Acked-by: Alex Deucher <alexander.deucher at amd.com>
> >
> > But I think maybe we should just drop all of this in the longer term.
> > We are slowly dropping every case where we reset the GPU.  I'm not
> > even sure it actually fixes the issue it was originally added to fix
> > at this point because the actual reset has moved to the no-irq
> > callback which will most likely not get called on an aborted suspend
> > anyway.
> >
>
> So perhaps for now this patch and in the next cycle or two maybe folks who worked
> on the aborted suspend case can try to trigger an aborted suspend w/ dGPUs + S3
> and see whether it's actually working and tear everything out if it doesn't as you say.
>

Yes, that was my thinking.

Alex


> > Alex
> >
> > > ---
> > >  drivers/gpu/drm/amd/amdgpu/amdgpu.h      |  1 +
> > >  drivers/gpu/drm/amd/amdgpu/amdgpu_acpi.c | 14 ++++++++++++++
> > >  drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c  |  2 +-
> > >  3 files changed, 16 insertions(+), 1 deletion(-)
> > >
> > > diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu.h
> > b/drivers/gpu/drm/amd/amdgpu/amdgpu.h
> > > index 3c20c2eadf4e..9cf3d65f17d7 100644
> > > --- a/drivers/gpu/drm/amd/amdgpu/amdgpu.h
> > > +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu.h
> > > @@ -1378,6 +1378,7 @@ static inline int
> > amdgpu_acpi_smart_shift_update(struct drm_device *dev,
> > >
> > >  #if defined(CONFIG_ACPI) && defined(CONFIG_SUSPEND)
> > >  bool amdgpu_acpi_is_s3_active(struct amdgpu_device *adev);
> > > +bool amdgpu_acpi_should_gpu_reset(struct amdgpu_device *adev);
> > >  bool amdgpu_acpi_is_s0ix_active(struct amdgpu_device *adev);
> > >  #else
> > >  static inline bool amdgpu_acpi_is_s0ix_active(struct amdgpu_device
> > *adev) { return false; }
> > > diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_acpi.c
> > b/drivers/gpu/drm/amd/amdgpu/amdgpu_acpi.c
> > > index 0e12315fa0cb..98ac53ee6bb5 100644
> > > --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_acpi.c
> > > +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_acpi.c
> > > @@ -1045,6 +1045,20 @@ bool amdgpu_acpi_is_s3_active(struct
> > amdgpu_device *adev)
> > >                 (pm_suspend_target_state == PM_SUSPEND_MEM);
> > >  }
> > >
> > > +/**
> > > + * amdgpu_acpi_should_gpu_reset
> > > + *
> > > + * @adev: amdgpu_device_pointer
> > > + *
> > > + * returns true if should reset GPU, false if not
> > > + */
> > > +bool amdgpu_acpi_should_gpu_reset(struct amdgpu_device *adev)
> > > +{
> > > +       if (adev->flags & AMD_IS_APU)
> > > +               return false;
> > > +       return pm_suspend_target_state != PM_SUSPEND_TO_IDLE;
> > > +}
> > > +
> > >  /**
> > >   * amdgpu_acpi_is_s0ix_active
> > >   *
> > > diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c
> > b/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c
> > > index 16871baee784..a84766c13ac5 100644
> > > --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c
> > > +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c
> > > @@ -2315,7 +2315,7 @@ static int amdgpu_pmops_suspend_noirq(struct
> > device *dev)
> > >         struct drm_device *drm_dev = dev_get_drvdata(dev);
> > >         struct amdgpu_device *adev = drm_to_adev(drm_dev);
> > >
> > > -       if (!adev->in_s0ix)
> > > +       if (amdgpu_acpi_should_gpu_reset(adev))
> > >                 return amdgpu_asic_reset(adev);
> > >
> > >         return 0;
> > > --
> > > 2.34.1
> > >