On Thu, 13 Jul 2023 at 11:41, Konrad Dybcio <konrad.dybcio@xxxxxxxxxx> wrote:
>
> On 12.07.2023 14:11, Dmitry Baryshkov wrote:
> > Add support for the MDSS cfg-cpu bus vote on the SM8450 platform.
> >
> > Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@xxxxxxxxxx>
> > ---
> > arch/arm64/boot/dts/qcom/sm8450.dtsi | 9 +++++++--
> > 1 file changed, 7 insertions(+), 2 deletions(-)
> >
> > diff --git a/arch/arm64/boot/dts/qcom/sm8450.dtsi b/arch/arm64/boot/dts/qcom/sm8450.dtsi
> > index 595533aeafc4..0b01f3027ee3 100644
> > --- a/arch/arm64/boot/dts/qcom/sm8450.dtsi
> > +++ b/arch/arm64/boot/dts/qcom/sm8450.dtsi
> > @@ -13,6 +13,7 @@
> > #include <dt-bindings/mailbox/qcom-ipcc.h>
> > #include <dt-bindings/phy/phy-qcom-qmp.h>
> > #include <dt-bindings/power/qcom-rpmpd.h>
> > +#include <dt-bindings/interconnect/qcom,icc.h>
> > #include <dt-bindings/interconnect/qcom,sm8450.h>
> > #include <dt-bindings/soc/qcom,gpr.h>
> > #include <dt-bindings/soc/qcom,rpmh-rsc.h>
> > @@ -2672,8 +2673,12 @@ mdss: display-subsystem@ae00000 {
> >
> > /* same path used twice */
> > interconnects = <&mmss_noc MASTER_MDP_DISP 0 &mc_virt SLAVE_EBI1_DISP 0>,
> > - <&mmss_noc MASTER_MDP_DISP 0 &mc_virt SLAVE_EBI1_DISP 0>;
> > - interconnect-names = "mdp0-mem", "mdp1-mem";
> > + <&mmss_noc MASTER_MDP_DISP 0 &mc_virt SLAVE_EBI1_DISP 0>,
> > + <&gem_noc MASTER_APPSS_PROC QCOM_ICC_TAG_ACTIVE_ONLY
> > + &config_noc SLAVE_DISPLAY_CFG QCOM_ICC_TAG_ACTIVE_ONLY>;
> Looking at icc_set_tag occurences in msm-5.10/techpack/display,
> I *think* active-only is only possible for the data bus (MDP-EBI)
Here I followed the vendor mdss fbdev driver (mdss_mdp.c), which
explicitly states:
static struct msm_bus_scale_pdata mdp_reg_bus_scale_table = {
.usecase = mdp_reg_bus_usecases,
.num_usecases = ARRAY_SIZE(mdp_reg_bus_usecases),
.name = "mdss_reg",
.active_only = true,
};
>
> Moreover, I think Linux is supposed to cast MDSS votes through the
> APPS RSC (so, nodes without _DISP [1][2]) and conversely, DISP_RSC is
> supposed to active-only votes
We can change this once your DISP_RSC lands. Anyway, I think we will
have to add the LLCC-MEM vote at some point later.
>
> Konrad
>
> [1] not that it matters today because it's not implemented yet
> [2] https://lore.kernel.org/linux-arm-msm/20230708-topic-rpmh_icc_rsc-v1-0-b223bd2ac8dd@xxxxxxxxxx
>
> > + interconnect-names = "mdp0-mem",
> > + "mdp1-mem",
> > + "cpu-cfg";
> >
> > resets = <&dispcc DISP_CC_MDSS_CORE_BCR>;
> >
--
With best wishes
Dmitry
