On 12-04-22, 09:34, Vladimir Zapolskiy wrote:
> Hi Bjorn,
>
> On 4/12/22 05:35, Bjorn Andersson wrote:
> > On Tue 05 Apr 12:38 CDT 2022, Dmitry Baryshkov wrote:
> >
> > > On Tue, 5 Apr 2022 at 20:17, Bjorn Andersson <bjorn.andersson@xxxxxxxxxx> wrote:
> > > >
> > > > On Tue 05 Apr 08:38 PDT 2022, Dmitry Baryshkov wrote:
> > > >
> > > > > On 11/03/2022 01:19, Vladimir Zapolskiy wrote:
> > > > > > All interconnect device tree nodes on sm8450 are 2-cells, however in
> > > > > > UFS node they are handled as 1-cells, fix it.
> > > > > >
> > > > > > Fixes: aa2d0bf04a3c ("arm64: dts: qcom: sm8450: add interconnect nodes")
> > > > > > Signed-off-by: Vladimir Zapolskiy <vladimir.zapolskiy@xxxxxxxxxx>
> > > > >
> > > > > Reviewed-by: Dmitry Baryshkov <dmitry.baryshkov@xxxxxxxxxx>
> > > > >
> > > > > Bjorn, could you please this pick for the -rc kernel?
> > > > >
> > > >
> > > > The change is obviously correct, but what difference does this change
> > > > make with the current implementation?
> > >
> > > it makes interconnect paths probe correctly. All NoC have
> > > #interconnec-cells = <2> now.
> > >
> >
> > But there's no code in the UFS driver that calls of_icc_get(), so what
> > does this actually do? (Other than correcting the dtb for the day when
> > we add that support to the driver).
>
> FWIW the change also has a runtime effect, it fixes a parsing of the board dtb,
> otherwise a warning in the kernel log appears:
>
> OF: /soc@0/ufshc@1d84000: could not get #interconnect-cells for /clocks/sleep-clk
>
> Why /clocks/sleep-clk is mentioned here at all??
> Its phandle value is 0x26, which is equal to SLAVE_UFS_MEM_CFG from the array.
We should either apply this fix or a patch to drop this line from dts.
Either would be apt and latter would make more sense..
--
~Vinod
