On 18.04.2024 1:49 AM, Dmitry Baryshkov wrote:
> On Wed, Apr 17, 2024 at 10:02:58PM +0200, Konrad Dybcio wrote:
>> There is no need to reinvent the wheel for simple read-match-set logic.
>>
>> Make speedbin discovery and assignment generation independent.
>>
>> This implicitly removes the bogus 0x80 / BIT(7) speed bin on A5xx,
>> which has no representation in hardware whatshowever.
>>
>> Signed-off-by: Konrad Dybcio <konrad.dybcio@xxxxxxxxxx>
>> ---
[...]
>> + /* No speedbins defined for this GPU SKU => allow all defined OPPs */
>> + if (!info->speedbins) {
>> + adreno_gpu->speedbin = ADRENO_SPEEDBIN_FUSE_NODATA;
>> + return devm_pm_opp_set_supported_hw(dev, &supp_hw, 1);
>
> BIT(0)
You mean for &supp_hw, or "1"?
1 is the "count" parameter, supp_hw is a "u32 supported_hw[count]"
>
>> + }
>> +
>> + /*
>> + * If a real error (not counting older devicetrees having no nvmem references)
>> + * occurs when trying to get the fuse value, bail out.
>> + */
>> + ret = adreno_read_speedbin(adreno_gpu, dev, &fuse);
>> + if (ret) {
>> + return ret;
>> + } else if (fuse == ADRENO_SPEEDBIN_FUSE_NODATA) {
>> + /* The info struct has speedbin data, but the DT is too old => allow all OPPs */
>
> Missing assignment to adeno_gpu->speedbin ? Or is it fine?
Good catch. Only mesa (and I suppose you :D) read this value.
>
>> + DRM_DEV_INFO(dev, "No GPU speed bin fuse, please update your device tree\n");
>> + return devm_pm_opp_set_supported_hw(dev, &supp_hw, 1);
>
> BIT(0)
> maybe #define it?
(ditto)
Konrad
