On Wed 04 Sep 03:20 PDT 2019, Brian Masney wrote:
> On Tue, Sep 03, 2019 at 10:01:03PM -0700, Bjorn Andersson wrote:
> > On Mon 02 Sep 14:19 PDT 2019, Brian Masney wrote:
> > > + mmssnoc: interconnect@fc478000 {
> > > + reg = <0xfc478000 0x4000>;
> > > + compatible = "qcom,msm8974-mmssnoc";
> > > + #interconnect-cells = <1>;
> > > + clock-names = "bus", "bus_a";
> > > + clocks = <&rpmcc RPM_SMD_MMSSNOC_AHB_CLK>,
> > > + <&rpmcc RPM_SMD_MMSSNOC_AHB_A_CLK>;
> >
> > Isn't MMSS_S0_AXI_CLK the bus clock of the mmssnoc (which somehow seems
> > to depend on mmssnoc_ahb_clk)?
>
> I'll give that a try. Do you know which clock I should use for bus_a
> here? On the mmcc, I see the following mmss clocks available:
>
> MMSS_AHB_CLK_SRC
> MMSS_AXI_CLK_SRC
> MMSS_RBCPR_CLK_SRC
> MMSS_MISC_AHB_CLK
> MMSS_MMSSNOC_AHB_CLK
> MMSS_MMSSNOC_BTO_AHB_CLK
> MMSS_MMSSNOC_AXI_CLK
> MMSS_S0_AXI_CLK
>
> I'm also unsure of what's going on at the hardware level that the second
> clock (bus_a) is needed.
>
In msm-3.4 clock-8974.c both bus and bus_a is defined as
mmss_s0_axi_clk. But iirc I also needed mmssnoc_axi_clk to get DSI
working on my devices, which is listed as .depends of the s0_axi clock.
So that probably needs some more investigation...But easiest would
probably be to just have the mmss_noc use them both listed as bus
clocks?
Regards,
Bjorn
> > > + mdss: mdss@fd900000 {
> >
> > I think you can omit the client, as this adheres to the standard binding
> > for interconnect clients. And you don't need to have an example that
> > covers all compatibles either...
>
> OK, I'll drop some of these.
>
> > > diff --git a/include/dt-bindings/interconnect/qcom,msm8974.h b/include/dt-bindings/interconnect/qcom,msm8974.h
> > > new file mode 100644
> > > index 000000000000..58acf7196410
> > > --- /dev/null
> > > +++ b/include/dt-bindings/interconnect/qcom,msm8974.h
> > > @@ -0,0 +1,146 @@
> > > +/* SPDX-License-Identifier: GPL-2.0 */
> >
> > Would you mind dual licensing this part as well?
>
> Sure, that was an oversight on my part.
>
> > Apart from that, I think this binding looks good.
>
> Thanks,
>
> Brian
