Re: [PATCH] arm64: dts: imx8mp: Add dsp node

On Fri, Jul 16, 2021 at 5:05 AM Peng Fan [off-list ref] wrote:

Hi Daniel

quoted

Subject: [PATCH] arm64: dts: imx8mp: Add dsp node

From: Daniel Baluta <daniel.baluta@nxp.com>

i.MX8 MPlus SoC integrates Cadence HIFI4 DSP. This core runs either a custom
firmware or the open source SOF firmware [1]

DSP device is handled by SOF OF driver found in sound/soc/sof/sof-of-dev.c

Notice that the DSP node makes use of:
      - dsp_reserved, a reserved memory region for various Audio
        resources (e.g firmware loading, audio buffers, etc).
      - Messaging Unit (mu2) for passing notifications betweem ARM
        core and DSP.

[1] https://thesofproject.github.io/latest/platforms/index.html
Signed-off-by: Daniel Baluta <daniel.baluta@nxp.com>
---
 arch/arm64/boot/dts/freescale/imx8mp.dtsi | 28
+++++++++++++++++++++++
 1 file changed, 28 insertions(+)

diff --git a/arch/arm64/boot/dts/freescale/imx8mp.dtsi

b/arch/arm64/boot/dts/freescale/imx8mp.dtsi
index 9f7c7f587d38..264ba540f363 100644

--- a/arch/arm64/boot/dts/freescale/imx8mp.dtsi
+++ b/arch/arm64/boot/dts/freescale/imx8mp.dtsi

@@ -135,6 +135,17 @@ clk_ext4: clock-ext4 {
              clock-output-names = "clk_ext4";
      };

+     reserved-memory {
+             #address-cells = <2>;
+             #size-cells = <2>;
+             ranges;
+
+             dsp_reserved: dsp@92400000 {
+                     reg = <0 0x92400000 0 0x2000000>;
+                     no-map;
+             };
+     };
+
      pmu {
              compatible = "arm,cortex-a53-pmu";
              interrupts = <GIC_PPI 7

@@ -698,6 +709,13 @@ mu: mailbox@30aa0000 {
                              #mbox-cells = <2>;
                      };

+                     mu2: mailbox@30e60000 {
+                             compatible = "fsl,imx8mp-mu", "fsl,imx6sx-mu";
+                             reg = <0x30e60000 0x10000>;
+                             interrupts = <GIC_SPI 136 IRQ_TYPE_LEVEL_HIGH>;
+                             #mbox-cells = <2>;

I think better mark this node as disabled. If its clock not enabled,
kernel may hang.

Good point. Fixed in v2.