Hello! I've run into an interesting problem with RPMsg communication between the Cortex-A35 and M4 cores on the IMX8QXP platform, and I could use some insights.
I'm working with Linux 6.6, and here's the situation: When I try to load the demo provided imx8qx_m4_TCM_rpmsg_lite_str_echo_rtos binary from U-boot, the system immediately restarts. Interestingly, the same setup works perfectly fine on Linux 5.15, which suggests something might have changed in the newer kernel version.
Here's my device tree configuration for reference:
/ {
reserved-memory {
#address-cells = <2>;
#size-cells = <2>;
ranges;
gpu_reserved: gpu_reserved@B0000000 {
no-map;
reg = <0 0xB0000000 0 0x08000000>;
};
vdev0vring0: vdev0vring0@90000000 {
reg = <0 0x90000000 0 0x8000>;
no-map;
};
vdev0vring1: vdev0vring1@90008000 {
reg = <0 0x90008000 0 0x8000>;
no-map;
};
vdev1vring0: vdev1vring0@90010000 {
reg = <0 0x90010000 0 0x8000>;
no-map;
};
vdev1vring1: vdev1vring1@90018000 {
reg = <0 0x90018000 0 0x8000>;
no-map;
};
rsc_table: rsc_table@900ff000 {
reg = <0 0x900ff000 0 0x1000>;
no-map;
};
vdevbuffer: vdevbuffer {
compatible = "shared-dma-pool";
reg = <0 0x90400000 0 0x100000>;
no-map;
};
dsp_reserved: dsp@92400000 {
reg = <0 0x92400000 0 0x1000000>;
no-map;
};
dsp_reserved_heap: dsp_reserved_heap {
reg = <0 0x93400000 0 0xef0000>;
no-map;
};
dsp_vdev0vring0: vdev0vring0@942f0000 {
reg = <0 0x942f0000 0 0x8000>;
no-map;
};
dsp_vdev0vring1: vdev0vring1@942f8000 {
reg = <0 0x942f8000 0 0x8000>;
no-map;
};
dsp_vdev0buffer: vdev0buffer@94300000 {
compatible = "shared-dma-pool";
reg = <0 0x94300000 0 0x100000>;
no-map;
};
/* global autoconfigured region for contiguous allocations */
linux,cma {
compatible = "shared-dma-pool";
reusable;
size = <0 0x08000000>;
alloc-ranges = <0 0xa8000000 0 0x08000000>;
linux,cma-default;
};
};
imx8x_cm4: imx8x_cm4@0 {
compatible = "fsl,imx8qxp-cm4";
rsc-da = <0x90000000>;
mbox-names = "tx", "rx", "rxdb";
mboxes = <&lsio_mu5 0 1
&lsio_mu5 1 1
&lsio_mu5 3 1>;
memory-region = <&vdevbuffer>, <&vdev0vring0>, <&vdev0vring1>,
<&vdev1vring0>, <&vdev1vring1>, <&rsc_table>;
fsl,resource-id = <IMX_SC_R_M4_0_PID0>;
fsl,entry-address = <0x34fe0000>;
status = "okay";
power-domains = <&pd IMX_SC_R_M4_0_PID0>,
<&pd IMX_SC_R_M4_0_MU_1A>;
};
rpmsg: rpmsg@0 {
compatible = "fsl,imx8qxp-rpmsg";
mbox-names = "tx", "rx", "rxdb";
mboxes = <&lsio_mu5 0 1
&lsio_mu5 1 1
&lsio_mu5 3 1>;
memory-region = <&vdevbuffer>;
mub-partition = <3>;
status = "okay";
};
};
&lsio_mu5 {
status = "okay";
};
My bootloader setup is straightforward:
setenv m4_0_image "imx8qx_m4_TCM_rpmsg_lite_str_echo_rtos.bin"
run m4boot_0
Has anyone encountered similar behaviour or have suggestions on what might be causing this regression in Linux 6.6? I'm particularly interested in:
- Changes in RPMsg handling between 5.15 and 6.6
- Potential memory mapping issues
- MU (Message Unit) configuration differences
Any insights or debugging suggestions would be greatly appreciated!
P.S : I noticed even if I run the sample demo imx8qx_m4_TCM_hello_world.bin, I got the same issue.
