On Wed, Oct 8, 2025 at 11:33 AM Mario Limonciello
<[email protected]> wrote:
>
> On 10/8/25 10:31 AM, Alex Deucher wrote:
> > On Thu, Oct 2, 2025 at 1:51 PM Mario Limonciello
> > <[email protected]> wrote:
> >>
> >> [Why]
> >> The shutdown() callback uses amdgpu_ip_suspend() which doesn't notify
> >> drm clients during shutdown.  This could lead to hangs.
> >>
> >> [How]
> >> Change amdgpu_pci_shutdown() to call the same sequence as suspend/resume.
> >>
> >> Signed-off-by: Mario Limonciello <[email protected]>
> >> ---
> >>   drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c | 3 ++-
> >>   1 file changed, 2 insertions(+), 1 deletion(-)
> >>
> >> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c 
> >> b/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c
> >> index 4905efa63ddc..e377fbb43e96 100644
> >> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c
> >> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c
> >> @@ -2558,7 +2558,8 @@ amdgpu_pci_shutdown(struct pci_dev *pdev)
> >>           */
> >>          if (!amdgpu_passthrough(adev))
> >>                  adev->mp1_state = PP_MP1_STATE_UNLOAD;
> >> -       amdgpu_device_ip_suspend(adev);
> >> +       amdgpu_device_prepare(dev);
> >> +       amdgpu_device_suspend(dev, true);
> >
> > I think we did this before, but users complained that it took longer
> > to shut down their systems.
>
> That's presumably from the evictions.  I think that it should be
> improved because we now have:
>
> drm/amd: Avoid evicting resources at S5

That makes sense.  The series is:
Reviewed-by: Alex Deucher <[email protected]>


>
> >
> > Alex
> >
> >>          adev->mp1_state = PP_MP1_STATE_NONE;
> >>   }
> >>
> >> --
> >> 2.51.0
> >>
>

Reply via email to