On Mon, Dec 16, 2024 at 01:11:35PM -0800, Abhinav Kumar wrote:
>
>
> On 12/16/2024 12:27 AM, Dmitry Baryshkov wrote:
> > Link DSPP_2 to the LM_2 and DSPP_3 to the LM_3 mixer blocks.
> >
> > Fixes: 05ae91d960fd ("drm/msm/dpu: enable DSPP support on SM8[12]50")
> > Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@xxxxxxxxxx>
> > ---
> > drivers/gpu/drm/msm/disp/dpu1/catalog/dpu_5_0_sm8150.h | 2 ++
> > 1 file changed, 2 insertions(+)
> >
>
> Change looks fine
>
> Reviewed-by: Abhinav Kumar <quic_abhinavk@xxxxxxxxxxx>
>
> One question below (not tied to the change but arose due to it):
>
> > diff --git a/drivers/gpu/drm/msm/disp/dpu1/catalog/dpu_5_0_sm8150.h b/drivers/gpu/drm/msm/disp/dpu1/catalog/dpu_5_0_sm8150.h
> > index 6ccfde82fecdb4e3612df161814b16f7af40ca5f..421afacb7248039abd9fb66bcb73b756ae0d640a 100644
> > --- a/drivers/gpu/drm/msm/disp/dpu1/catalog/dpu_5_0_sm8150.h
> > +++ b/drivers/gpu/drm/msm/disp/dpu1/catalog/dpu_5_0_sm8150.h
> > @@ -164,6 +164,7 @@ static const struct dpu_lm_cfg sm8150_lm[] = {
> > .sblk = &sdm845_lm_sblk,
> > .lm_pair = LM_3,
> > .pingpong = PINGPONG_2,
> > + .dspp = DSPP_2,
> > }, {
> > .name = "lm_3", .id = LM_3,
> > .base = 0x47000, .len = 0x320,
> > @@ -171,6 +172,7 @@ static const struct dpu_lm_cfg sm8150_lm[] = {
> > .sblk = &sdm845_lm_sblk,
> > .lm_pair = LM_2,
> > .pingpong = PINGPONG_3,
> > + .dspp = DSPP_3,
> > }, {
> > .name = "lm_4", .id = LM_4,
> > .base = 0x48000, .len = 0x320,
> >
>
> the consumer of .dspp seems to be in the RM code which is used to map the
> DSPP to encoder_id but is there really any case where lm_id != dspp_id ... I
> guess I am missing the context of why DSPP id needs to be tracked as LMs and
> DSPPs go together. Let me also check this part internally.
For example check the SDM845, the LM_5 is tied to DSPP_3.
LM | DSPP
---------
0 | 0
1 | 1
2 | 2
5 | 3
--
With best wishes
Dmitry
