[PATCH] drm/i915/dp_mst: Fix dsc mst bw overhead calculation

Nautiyal, Ankit K ankit.k.nautiyal at intel.com
Wed Oct 9 11:10:41 UTC 2024


On 10/9/2024 4:15 PM, Kandpal, Suraj wrote:
>
>> -----Original Message-----
>> From: Jani Nikula <jani.nikula at linux.intel.com>
>> Sent: Wednesday, October 9, 2024 3:30 PM
>> To: Kandpal, Suraj <suraj.kandpal at intel.com>; intel-
>> xe at lists.freedesktop.org; intel-gfx at lists.freedesktop.org
>> Cc: Nautiyal, Ankit K <ankit.k.nautiyal at intel.com>; Kandpal, Suraj
>> <suraj.kandpal at intel.com>
>> Subject: Re: [PATCH] drm/i915/dp_mst: Fix dsc mst bw overhead calculation
>>
>> On Wed, 09 Oct 2024, Suraj Kandpal <suraj.kandpal at intel.com> wrote:
>>> Fix the DSC flag assignment based on the dsc_slice_count returned to
>>> avoid divide by zero error.
>>>
>>> Fixes: 4e0837a8d00a ("drm/i915/dp_mst: Account for FEC and DSC
>>> overhead during BW allocation")
>>> Signed-off-by: Suraj Kandpal <suraj.kandpal at intel.com>
>>> ---
>>>   drivers/gpu/drm/i915/display/intel_dp_mst.c | 7 ++++++-
>>>   1 file changed, 6 insertions(+), 1 deletion(-)
>>>
>>> diff --git a/drivers/gpu/drm/i915/display/intel_dp_mst.c
>>> b/drivers/gpu/drm/i915/display/intel_dp_mst.c
>>> index 4765bda154c1..bacd294d6bb6 100644
>>> --- a/drivers/gpu/drm/i915/display/intel_dp_mst.c
>>> +++ b/drivers/gpu/drm/i915/display/intel_dp_mst.c
>>> @@ -105,11 +105,16 @@ static int intel_dp_mst_bw_overhead(const
>> struct intel_crtc_state *crtc_state,
>>>   	if (dsc) {
>>>   		int num_joined_pipes =
>> intel_crtc_num_joined_pipes(crtc_state);
>>> -		flags |= DRM_DP_BW_OVERHEAD_DSC;
>>>   		dsc_slice_count = intel_dp_dsc_get_slice_count(connector,
>>>   							       adjusted_mode-
>>> clock,
>>>   							       adjusted_mode-
>>> hdisplay,
>>>
>> num_joined_pipes);
>>> +		/*
>>> +		 * Try with dsc only if dsc_slice_count has a sane value i.e
>> value is no
>>> +		 * 0
>>> +		 */
>>> +		if (dsc_slice_count)
>>> +			flags |= DRM_DP_BW_OVERHEAD_DSC;
>> Do you think that's enough to handle the error?!
> Well this will make sure that if dsc_slice_count ends up being zero we don't take dsc overhead into account.
> Which should be enough to make sure we don't go and end up having a divide by zero error

But the overhead computed will not be correct. I was thinking to avoid 
setting dsc flag if we dont get a valid slice_count.

Perhaps in intel_dp_dsc_mst_compute_link_config() before calling 
intel_dp_mst_find_vcpi_slots_for_bpp() we should check for slice_count.

Actually we again compute dsc slice_count quite late in 
dp_dsc_compute_config and set that in pipe_config->dsc.slice_count .


Regards,

Ankit


>
> Regards,
> Suraj Kandpal
>> BR,
>> Jani.
>>
>>>   	}
>>>
>>>   	overhead = drm_dp_bw_overhead(crtc_state->lane_count,
>> --
>> Jani Nikula, Intel


More information about the Intel-xe mailing list