Re: [PATCH 5/7] drm/msm/dpu: inline dpu_encoder_get_wb()

[Dmitry Baryshkov <dmitry.baryshkov@linaro.org>]

On 03/05/2023 02:58, Abhinav Kumar wrote:
> 
> 
> On 5/2/2023 4:54 PM, Dmitry Baryshkov wrote:
>> On 03/05/2023 02:51, Abhinav Kumar wrote:
>>>
>>>
>>> On 4/30/2023 4:57 PM, Dmitry Baryshkov wrote:
>>>> The function dpu_encoder_get_wb() returns controller_id if the
>>>> corresponding WB is present in the catalog. We can inline this function
>>>> and rely on dpu_rm_get_wb() returning NULL for indices for which the
>>>> WB is not present on the device.
>>>>
>>>> Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
>>>> ---
>>>>   drivers/gpu/drm/msm/disp/dpu1/dpu_encoder.c | 24 
>>>> ++-------------------
>>>>   1 file changed, 2 insertions(+), 22 deletions(-)
>>>>
>>>> diff --git a/drivers/gpu/drm/msm/disp/dpu1/dpu_encoder.c 
>>>> b/drivers/gpu/drm/msm/disp/dpu1/dpu_encoder.c
>>>> index 4c85cbb030e4..507ff3f88c67 100644
>>>> --- a/drivers/gpu/drm/msm/disp/dpu1/dpu_encoder.c
>>>> +++ b/drivers/gpu/drm/msm/disp/dpu1/dpu_encoder.c
>>>> @@ -1277,22 +1277,6 @@ static enum dpu_intf 
>>>> dpu_encoder_get_intf(const struct dpu_mdss_cfg *catalog,
>>>>       return INTF_MAX;
>>>>   }
>>>> -static enum dpu_wb dpu_encoder_get_wb(const struct dpu_mdss_cfg 
>>>> *catalog,
>>>> -        enum dpu_intf_type type, u32 controller_id)
>>>> -{
>>>> -    int i = 0;
>>>> -
>>>> -    if (type != INTF_WB)
>>>> -        return WB_MAX;
>>>> -
>>>> -    for (i = 0; i < catalog->wb_count; i++) {
>>>> -        if (catalog->wb[i].id == controller_id)
>>>> -            return catalog->wb[i].id;
>>>> -    }
>>>> -
>>>> -    return WB_MAX;
>>>> -}
>>>> -
>>>>   void dpu_encoder_vblank_callback(struct drm_encoder *drm_enc,
>>>>           struct dpu_encoder_phys *phy_enc)
>>>>   {
>>>> @@ -2261,7 +2245,6 @@ static int dpu_encoder_setup_display(struct 
>>>> dpu_encoder_virt *dpu_enc,
>>>>            */
>>>>           u32 controller_id = disp_info->h_tile_instance[i];
>>>>           enum dpu_intf intf_idx;
>>>> -        enum dpu_wb wb_idx;
>>>>           if (disp_info->num_of_h_tiles > 1) {
>>>>               if (i == 0)
>>>> @@ -2279,14 +2262,11 @@ static int dpu_encoder_setup_display(struct 
>>>> dpu_encoder_virt *dpu_enc,
>>>>                                   disp_info->intf_type,
>>>>                                   controller_id);
>>>> -        wb_idx = dpu_encoder_get_wb(dpu_kms->catalog,
>>>> -                disp_info->intf_type, controller_id);
>>>> -
>>>>           if (intf_idx >= INTF_0 && intf_idx < INTF_MAX)
>>>>               phys_params.hw_intf = dpu_rm_get_intf(&dpu_kms->rm, 
>>>> intf_idx);
>>>> -        if (wb_idx >= WB_0 && wb_idx < WB_MAX)
>>>> -            phys_params.hw_wb = dpu_rm_get_wb(&dpu_kms->rm, wb_idx);
>>>> +        if (disp_info->intf_type == INTF_WB && controller_id < WB_MAX)
>>>> +            phys_params.hw_wb = dpu_rm_get_wb(&dpu_kms->rm, 
>>>> controller_id);
>>>
>>>
>>>  From what I see, with 
>>> https://patchwork.freedesktop.org/patch/534776/?series=117146&rev=1 
>>> we are dropping those checks from the RM too, so we are going to rely 
>>> totally on entering the values correctly in catalog from now on?
>>
>> Yes. I see no reason to mistrust the kernel data itself.
> 
> Alright, if thats the overall plan, this change itself is fine.

Yes, I think this is what we discussed some time ago for UBWC and QSEED 
programming.

> 
> Reviewed-by: Abhinav Kumar <quic_abhinavk@quicinc.com>
>>
>>>
>>>>           if (!phys_params.hw_intf && !phys_params.hw_wb) {
>>>>               DPU_ERROR_ENC(dpu_enc, "no intf or wb block assigned 
>>>> at idx: %d\n", i);
>>

-- 
With best wishes
Dmitry