Re: [Intel-gfx] [PATCH v3 3/4] drm/i915: Perform TLB invalidation on all GTs during suspend/resume

[Jani Nikula <jani.nikula@linux.intel.com>]

On Tue, 03 Oct 2023, John Harrison <john.c.harrison@intel.com> wrote:
> On 10/3/2023 08:59, Andi Shyti wrote:
>> Hi Jani,
>>
>>>> Consider multi-gt support when cancelling all tlb invalidations on
>>>> suspend, and when submitting tlb invalidations on resume.
>>>>
>>>> Suggested-by: Tvrtko Ursulin <tvrtko.ursulin@linux.intel.com>
>>>> Signed-off-by: Fei Yang <fei.yang@intel.com>
>>>> Signed-off-by: Jonathan Cavitt <jonathan.cavitt@intel.com>
>>>> CC: John Harrison <John.C.Harrison@Intel.com>
>>> I guess I'm wondering why the top level suspend hook needs to iterate
>>> gts instead of some lower level thing. We should aim to reduce
>>> gem/gt/display details from the top level.
>> I'm not sure I am understanding the question.
>>
>> The TLB invalidation details are kept under the GT. But when
>> suspend is called, then the GT invalidation has to be triggered
>> by the top levels for each GT. Right?
> I think Jani's point is that the top level should be:
> i915_drm_suspend(...) {
>     ...
>     intel_tlb_suspend(dev_priv);
> }
>
> Then the TLB suspend helper function calls into the GT / UC layers as 
> appropriate. But none of that internal only detail is exposed at the top 
> level.

That's right, thanks for clarifying.

BR,
Jani.


>
> John.
>
>>
>> Thanks,
>> Andi
>

-- 
Jani Nikula, Intel