On Thu, 27 Jul 2023 at 01:30, Abhinav Kumar <quic_abhinavk@xxxxxxxxxxx> wrote:
>
>
>
> On 5/21/2023 10:10 AM, Dmitry Baryshkov wrote:
> > As we are going to use MDSS data for DPU programming, populate missing
> > MDSS data. The UBWC 1.0 and no UBWC cases do not require MDSS
> > programming, so skip them.
> >
>
> Can you pls point me to the downstream references you used for msm8998?
msm-3.18, drivers/video/msm/mdss/mdss_mdp.c
See the function mdss_mdp_hw_rev_caps_init(). It sets has_ubwc for MDP
1.07 (msm8996), 1.14 (msm8937) / 1.16 (msm8953) and 3.0 (msm8998).
> Was that just taken from catalog? If so I would ask for the reference
> for the catalog too.
>
> As per the register the decoder version is 0x0 and not 1.
>
> Even encoder version is 0 from what i see and not 1. Thats why a
> dec_version of UBWC_1_0 is not doing anything i assume.
>
> Some additional questions:
>
> 1) Does the whole chunk in dpu_hw_sspp_setup_format() which handles ubwc
> programming need to be protected by if (ctx->ubwc) now ?
It's hard to discuss the question which is irrelevant for this patch.
Nevertheless, yes, it needs to be protected because e.g. qcm2290
doesn't have UBWC support.
> 2) The values of UBWC_x_x dont match the values of DPU_HW_UBWC_VER_xx.
> What was the reason for the catalog to go with DPU_HW_UBWC_VER_xx in the
> catalog for the encoder version in the first place? Because looking at
> the registers UBWC_x_x is the correct value.
Huh. This definitely should be asked next to the code that you wish to
discuss. The DPU_HW_UBWC_VER_xx values come from the first DPU
revision.
>
>
> > Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@xxxxxxxxxx>
> > ---
> > drivers/gpu/drm/msm/msm_mdss.c | 21 +++++++++++++++++++--
> > 1 file changed, 19 insertions(+), 2 deletions(-)
> >
> > diff --git a/drivers/gpu/drm/msm/msm_mdss.c b/drivers/gpu/drm/msm/msm_mdss.c
> > index ed836c659688..9bb7be4b9ebb 100644
> > --- a/drivers/gpu/drm/msm/msm_mdss.c
> > +++ b/drivers/gpu/drm/msm/msm_mdss.c
> > @@ -264,6 +264,10 @@ static int msm_mdss_enable(struct msm_mdss *msm_mdss)
> > * UBWC_n and the rest of params comes from hw data.
> > */
> > switch (msm_mdss->mdss_data->ubwc_dec_version) {
> > + case 0: /* no UBWC */
> > + case UBWC_1_0:
> > + /* do nothing */
> > + break;
> > case UBWC_2_0:
> > msm_mdss_setup_ubwc_dec_20(msm_mdss);
> > break;
> > @@ -502,10 +506,22 @@ static int mdss_remove(struct platform_device *pdev)
> > return 0;
> > }
> >
> > +static const struct msm_mdss_data msm8998_data = {
> > + .ubwc_enc_version = UBWC_1_0,
> > + .ubwc_dec_version = UBWC_1_0,
> > + .highest_bank_bit = 1,
> > +};
> > +
> > +static const struct msm_mdss_data qcm2290_data = {
> > + /* no UBWC */
> > + .highest_bank_bit = 0x2,
> > +};
> > +
> > static const struct msm_mdss_data sc7180_data = {
> > .ubwc_enc_version = UBWC_2_0,
> > .ubwc_dec_version = UBWC_2_0,
> > .ubwc_static = 0x1e,
> > + .highest_bank_bit = 0x3,
> > };
> >
> > static const struct msm_mdss_data sc7280_data = {
> > @@ -550,6 +566,7 @@ static const struct msm_mdss_data sm6115_data = {
> > .ubwc_dec_version = UBWC_2_0,
> > .ubwc_swizzle = 7,
> > .ubwc_static = 0x11f,
> > + .highest_bank_bit = 0x1,
> > };
> >
> > static const struct msm_mdss_data sm8250_data = {
> > @@ -574,8 +591,8 @@ static const struct msm_mdss_data sm8550_data = {
> >
> > static const struct of_device_id mdss_dt_match[] = {
> > { .compatible = "qcom,mdss" },
> > - { .compatible = "qcom,msm8998-mdss" },
> > - { .compatible = "qcom,qcm2290-mdss" },
> > + { .compatible = "qcom,msm8998-mdss", .data = &msm8998_data },
> > + { .compatible = "qcom,qcm2290-mdss", .data = &qcm2290_data },
> > { .compatible = "qcom,sdm845-mdss", .data = &sdm845_data },
> > { .compatible = "qcom,sc7180-mdss", .data = &sc7180_data },
> > { .compatible = "qcom,sc7280-mdss", .data = &sc7280_data },
--
With best wishes
Dmitry
