Re: [PATCH] drm/xe: Revert "drm/i915: Disable DSB in Xe KMD"

[Maarten Lankhorst <maarten.lankhorst@linux.intel.com>]

Hey,

Den 2024-07-09 kl. 12:23, skrev Matthew Auld:
> On 09/07/2024 10:33, Maarten Lankhorst wrote:
>> Hello,
>>
>> I've been trying out this patch, but I'm still seeing corruption on my laptop with DSB enabled.
>> I'd recommend not merging until it is fixed. :)
> 
> Just wondering, is it correct to use cpu:wb for the dsb buffer? Does the dsb stuff not suffer from "display is not coherent with gpu/cpu caches"? In i915 it looks to be cpu:wc.

My guess is that could be it. Our kernel mappings are cached by default.

Perhaps I'll look at the DPT code too, it could suffer from the same issue.

Will test a bit more tomorrow :)

~Maarten

>> Cheers,
>> ~Maarten
>>
>> Den 2024-07-08 kl. 07:56, skrev Animesh Manna:
>>> This reverts commit c27f010aa1884276ee5dae72034d84987060c769.
>>>
>>> After fix from [1] dsb timeout issue is not reproducible on local testing
>>> with xe driver. Checking CI result to confirm and not for review.
>>>
>>> [1] https://patchwork.freedesktop.org/series/130783/
>>>
>>> Signed-off-by: Animesh Manna <animesh.manna@intel.com>
>>> ---
>>>   drivers/gpu/drm/i915/display/intel_dsb.c | 4 ----
>>>   1 file changed, 4 deletions(-)
>>>
>>> diff --git a/drivers/gpu/drm/i915/display/intel_dsb.c b/drivers/gpu/drm/i915/display/intel_dsb.c
>>> index 2ab3765f6c06..e9c4835a2c30 100644
>>> --- a/drivers/gpu/drm/i915/display/intel_dsb.c
>>> +++ b/drivers/gpu/drm/i915/display/intel_dsb.c
>>> @@ -463,10 +463,6 @@ struct intel_dsb *intel_dsb_prepare(struct intel_atomic_state *state,
>>>       if (!i915->display.params.enable_dsb)
>>>           return NULL;
>>>   -    /* TODO: DSB is broken in Xe KMD, so disabling it until fixed */
>>> -    if (!IS_ENABLED(I915))
>>> -        return NULL;
>>> -
>>>       dsb = kzalloc(sizeof(*dsb), GFP_KERNEL);
>>>       if (!dsb)
>>>           goto out;