| Message ID | 20201201042520.2116271-2-vkoul@kernel.org |
|---|---|
| State | New |
| Headers | show |
| Series | [v2,1/2] arm64: dts: qcom: sdm845: Add gpi dma node | expand |
On Mon 30 Nov 22:25 CST 2020, Vinod Koul wrote: > gpi_dma0 can be used for spi and i2c transfers on db845c, so enable it > > Signed-off-by: Vinod Koul <vkoul@kernel.org> > --- > arch/arm64/boot/dts/qcom/sdm845-db845c.dts | 4 ++++ > 1 file changed, 4 insertions(+) > > diff --git a/arch/arm64/boot/dts/qcom/sdm845-db845c.dts b/arch/arm64/boot/dts/qcom/sdm845-db845c.dts > index 7cc236575ee2..0653468f26ce 100644 > --- a/arch/arm64/boot/dts/qcom/sdm845-db845c.dts > +++ b/arch/arm64/boot/dts/qcom/sdm845-db845c.dts > @@ -418,6 +418,10 @@ &gcc { > <GCC_QSPI_CNOC_PERIPH_AHB_CLK>; > }; > > +&gpi_dma0 { > + status = "okay"; > +}; > + But you're only enabling the gpi, you don't have anything wired to it, so why enable it? I took a look downstream and they have all the i2c and spi nodes with their dmas filled out - and the two gpis enabled at all times (sdm845.dtsi). Can/should we do the same? Regards, Bjorn > &gpu { > zap-shader { > memory-region = <&gpu_mem>; > -- > 2.26.2 >
On 01-12-20, 19:06, Bjorn Andersson wrote: > On Mon 30 Nov 22:25 CST 2020, Vinod Koul wrote: > > > gpi_dma0 can be used for spi and i2c transfers on db845c, so enable it > > > > Signed-off-by: Vinod Koul <vkoul@kernel.org> > > --- > > arch/arm64/boot/dts/qcom/sdm845-db845c.dts | 4 ++++ > > 1 file changed, 4 insertions(+) > > > > diff --git a/arch/arm64/boot/dts/qcom/sdm845-db845c.dts b/arch/arm64/boot/dts/qcom/sdm845-db845c.dts > > index 7cc236575ee2..0653468f26ce 100644 > > --- a/arch/arm64/boot/dts/qcom/sdm845-db845c.dts > > +++ b/arch/arm64/boot/dts/qcom/sdm845-db845c.dts > > @@ -418,6 +418,10 @@ &gcc { > > <GCC_QSPI_CNOC_PERIPH_AHB_CLK>; > > }; > > > > +&gpi_dma0 { > > + status = "okay"; > > +}; > > + > > But you're only enabling the gpi, you don't have anything wired to it, > so why enable it? > > I took a look downstream and they have all the i2c and spi nodes with > their dmas filled out - and the two gpis enabled at all times > (sdm845.dtsi). Can/should we do the same? So dma is merged and in linux-next but support for spi/i2c is not as that is dependent upon dmaengine interface introduced for this. I was planning to post spi/i2c after rc1 and also the dts enabling them
diff --git a/arch/arm64/boot/dts/qcom/sdm845-db845c.dts b/arch/arm64/boot/dts/qcom/sdm845-db845c.dts index 7cc236575ee2..0653468f26ce 100644 --- a/arch/arm64/boot/dts/qcom/sdm845-db845c.dts +++ b/arch/arm64/boot/dts/qcom/sdm845-db845c.dts @@ -418,6 +418,10 @@ &gcc { <GCC_QSPI_CNOC_PERIPH_AHB_CLK>; }; +&gpi_dma0 { + status = "okay"; +}; + &gpu { zap-shader { memory-region = <&gpu_mem>;
gpi_dma0 can be used for spi and i2c transfers on db845c, so enable it Signed-off-by: Vinod Koul <vkoul@kernel.org> --- arch/arm64/boot/dts/qcom/sdm845-db845c.dts | 4 ++++ 1 file changed, 4 insertions(+)