Re: [PATCH V3 1/3] arm64: dts: sc7280: Add QSPI node

[Bjorn Andersson <bjorn.andersson@linaro.org>]

On Wed 14 Jul 02:47 CDT 2021, rojay@codeaurora.org wrote:

> On 2021-07-09 06:26, Stephen Boyd wrote:
> > Quoting rojay@codeaurora.org (2021-07-06 02:19:27)
> > > On 2021-06-08 13:37, rojay@codeaurora.org wrote:
> > > > On 2021-06-06 09:25, Bjorn Andersson wrote:
> > > >> On Fri 04 Jun 08:54 CDT 2021, Roja Rani Yarubandi wrote:
> > > >>
> > > >>> Add QSPI DT node for SC7280 SoC.
> > > >>>
> > > >>> Signed-off-by: Roja Rani Yarubandi <rojay@codeaurora.org>
> > > >>> ---
> > > >>> Changes in V3:
> > > >>>  - Broken the huge V2 patch into 3 smaller patches.
> > > >>>    1. QSPI DT nodes
> > > >>>    2. QUP wrapper_0 DT nodes
> > > >>>    3. QUP wrapper_1 DT nodes
> > > >>>
> > > >>> Changes in V2:
> > > >>>  - As per Doug's comments removed pinmux/pinconf subnodes.
> > > >>>  - As per Doug's comments split of SPI, UART nodes has been done.
> > > >>>  - Moved QSPI node before aps_smmu as per the order.
> > > >>>
> > > >>>  arch/arm64/boot/dts/qcom/sc7280-idp.dts | 29 ++++++++++++
> > > >>>  arch/arm64/boot/dts/qcom/sc7280.dtsi    | 61
> > > >>> +++++++++++++++++++++++++
> > > >>>  2 files changed, 90 insertions(+)
> > > >>>
> > > >>> diff --git a/arch/arm64/boot/dts/qcom/sc7280-idp.dts
> > > >>> b/arch/arm64/boot/dts/qcom/sc7280-idp.dts
> > > >>> index 3900cfc09562..d0edffc15736 100644
> > > >>> --- a/arch/arm64/boot/dts/qcom/sc7280-idp.dts
> > > >>> +++ b/arch/arm64/boot/dts/qcom/sc7280-idp.dts
> > > >>> @@ -268,6 +268,22 @@ pmr735b_die_temp {
> > > >>>             };
> > > >>>  };
> > > >>>
> > > >>> +&qspi {
> > > >>> +   status = "okay";
> > > >>> +   pinctrl-names = "default";
> > > >>> +   pinctrl-0 = <&qspi_clk>, <&qspi_cs0>, <&qspi_data01>;
> > > >>> +
> > > >>> +   flash@0 {
> > > >>> +           compatible = "jedec,spi-nor";
> > > >>> +           reg = <0>;
> > > >>> +
> > > >>> +           /* TODO: Increase frequency after testing */
> > > >>> +           spi-max-frequency = <25000000>;
> > > >>> +           spi-tx-bus-width = <2>;
> > > >>> +           spi-rx-bus-width = <2>;
> > > >>> +   };
> > > >>> +};
> > > >>> +
> > > >>>  &qupv3_id_0 {
> > > >>>     status = "okay";
> > > >>>  };
> > > >>> @@ -278,6 +294,19 @@ &uart5 {
> > > >>>
> > > >>>  /* PINCTRL - additions to nodes defined in sc7280.dtsi */
> > > >>>
> > > >>> +&qspi_cs0 {
> > > >>> +   bias-disable;
> > > >>> +};
> > > >>> +
> > > >>> +&qspi_clk {
> > > >>> +   bias-disable;
> > > >>> +};
> > > >>> +
> > > >>> +&qspi_data01 {
> > > >>> +   /* High-Z when no transfers; nice to park the lines */
> > > >>> +   bias-pull-up;
> > > >>> +};
> > > >>> +
> > > >>>  &qup_uart5_default {
> > > >>>     tx {
> > > >>>             pins = "gpio46";
> > > >>> diff --git a/arch/arm64/boot/dts/qcom/sc7280.dtsi
> > > >>> b/arch/arm64/boot/dts/qcom/sc7280.dtsi
> > > >>> index 6c9d5eb93f93..3047ab802cd2 100644
> > > >>> --- a/arch/arm64/boot/dts/qcom/sc7280.dtsi
> > > >>> +++ b/arch/arm64/boot/dts/qcom/sc7280.dtsi
> > > >>> @@ -1061,6 +1061,42 @@ apss_merge_funnel_in: endpoint {
> > > >>>                     };
> > > >>>             };
> > > >>>
> > > >>> +           qspi_opp_table: qspi-opp-table {
> > > >>
> > > >> This node doesn't represents anything on the mmio bus, so it shouldn't
> > > >> live in in /soc. Can't you move it into &qspi?
> > > >>
> > > >> Regards,
> > > >> Bjorn
> > > >>
> > > >
> > > > Sure, will move it into qspi node.
> > > >
> > > > Thanks,
> > > > Roja
> > > >
> > > 
> > > Hi Bjorn,
> > > 
> > > Moving "qspi_opp_table" inside &qspi node causing this warning:
> > > arch/arm64/boot/dts/qcom/sc7280.dtsi:1055.35-1072.6: Warning
> > > (spi_bus_reg): /soc@0/spi@88dc000/qspi-opp-table: missing or empty reg
> > > property
> > 
> > If DT folks are OK I think we should hard-code 'opp-table' as not a
> > device for spi to populate on the spi bus and relax the warning in the
> > devicetree compiler (see [1] for more details). Technically, nodes that
> > are bus controllers assume all child nodes are devices on that bus.  In
> > this case, we want to stick the opp table as a child of the spi node so
> > that it can be called 'opp-table' and not be a node in the root of DT.
> > 
> > > 
> > > Shall I keep the qspi-opp-table out of &qspi node?
> > > 
> > 
> > If you do, please move it to / instead of putting it under /soc as it
> > doesn't have an address or a reg property.
> > 
> 
> Hi Bjorn, Rob
> 
> Can we move this "qspi_opp_table" to / from /soc?
> 

If you have made a proper attempt to convince Rob and Mark that
a child "opp-table" in a SPI master is not a SPI device - and the
conclusion is that this is not a good idea...then yes it should live
outside /soc.

Thanks,
Bjorn