How to build the legacy_apps for run on linux a72 in vek280 board?

[yybingyybing]

Hello,
We can complie the legacy_apps follow the doc https://github.com/OpenAMP/openamp-system-reference/blob/main/examples/legacy_apps/README.md. But how could we build the app to run in vek280? Dose it same as the building in open-amp library?

Thanks.

[tnmysh]

@yybingyybing Thanks for your question.

Building apps for Xilinx Boards (specially for cortex-R) is bit more complicated. Some more libraries are needed that uses Xilinx BSP APIs. We usually use Vitis (IDE provided by AMD-Xilinx) to build these demos. Vitis internally builds Xilinx BSP libraries and link with these demos.

Please refer to UG1186 document below:

https://docs.amd.com/r/en-US/ug1186-zynq-openamp-gsg/Building-Remote-Applications-in-AMD-Vitis

This section: Building Remote Applications in AMD Vitis

Thanks.

[yybingyybing]

@yybingyybing Thanks for your question.

Building apps for Xilinx Boards (specially for cortex-R) is bit more complicated. Some more libraries are needed that uses Xilinx BSP APIs. We usually use Vitis (IDE provided by AMD-Xilinx) to build these demos. Vitis internally builds Xilinx BSP libraries and link with these demos.

Please refer to UG1186 document below:

https://docs.amd.com/r/en-US/ug1186-zynq-openamp-gsg/Building-Remote-Applications-in-AMD-Vitis

This section: Building Remote Applications in AMD Vitis

Thanks.

Helllo @tnmysh , thanks for your reply.

We could complise the remote app follow ug1186, but now want to build the ehco_test app run in linux, These is nothing about this in the ug1186. We use the kernel 6.12 in petalinux 2025.1 .
When we ran the echo_test which in /usr/bin/ , we got message:
No dev file for rpmsg-openamp-demo-channel in /sys/bus/rpmsg/devices
access(/sys/bus/rpmsg/devices/virtio0.rpmsg-openamp-demo-channel.-1.0): No such file or directory

Now, I find the freertos app remote using maybe has some errors, it seem had not created the rpmsg channel rpmsg-openamp-demo-channel.-1.0 . But I can't find the reason for this problem. When I use jtag run the remote app, I had not get any error message.

Another question is how can we use the trace0 in linux, defult it has nothing output.

[tnmysh]

HI @yybingyybing

There are two apps:

1. echo_test, and 2) rpmsg-echo-ping

2. echo_test -> This is using RPMsg implementation from kernelspace.

For this app to work, you need to mention shared memory, mailbox nodes under remoteproc nodes.

Please refer to this bindings in your kernel:

https://git.kernel.org/pub/scm/linux/kernel/git/remoteproc/linux.git/tree/Documentation/devicetree/bindings/remoteproc/xlnx,zynqmp-r5fss.yaml?h=for-next

The app can be compiled using this Makefile:
https://github.com/OpenAMP/openamp-system-reference/blob/main/examples/linux/rpmsg-echo-test/Makefile
app source code: https://github.com/OpenAMP/openamp-system-reference/blob/main/examples/linux/rpmsg-echo-test/echo_test.c

This is preferred approach.

2. rpmsg-echo-ping -> This is using RPMsg implementation from open-amp library built for linux userspace.

For this application you need the device-tree you mentioned above. It uses linux kernel UIO framework to access shared memory and IPI (Inter Processor Interrupt) address space.
This is not approach of RPMsg communication is not maintained very well and so not preferred.

The source code of this example is here: https://github.com/OpenAMP/openamp-system-reference/blob/main/examples/legacy_apps/examples/echo/rpmsg-ping.c

I think the problem here is, you are trying to use echo_test (approach-1) with wrong type of device-tree.

Edit: source code link of 2) rpmsg-echo-ping app.

[yybingyybing]

HI @tnmysh
Thanks for you reply again.

We only want to test the data exchange between linux A72 and freertos r5, no matter what mode to use .
Now we could use approach-1, the device-tree is configure as the xlnx,zynqmp-r5fss.yaml?h=for-next show. But there are still some issues. When we loaded the r5 elf program, we only got the device node virtio0.rpmsg_ctrl.0.0 virtio0.rpmsg_ns.53.53 in /sys/bus/rpmsg/devices/ , and had not virtio0.rpmsg-openamp-demo-channel.-1.0 , so the echo_test get 'No dev file for rpmsg-openamp-demo-channel in /sys/bus/rpmsg/devices ' errors.
I don't konw how to fix this problem. It looks like the r5 firmware has some errors, but we create the r5 firmware use vitis example.

I compared the running logs of zcu102 and found that the rpmsg channel is created after loading the R5 firmware, so I initially thought there was a problem with the R5 firmware. However, now I cannot see the running logs of R5 on my board, so I don't know where the error occurred. When I cat the trace0 in remote0 nodes, I get nothing. Does the trace0 node dont support cat command?

`
remoteproc@ffe00000 {
compatible = "xlnx,versal-r5fss";
xlnx,cluster-mode = <0x00>;
xlnx,tcm-mode = <0x00>;
#address-cells = <0x02>;
#size-cells = <0x02>;
ranges = <0x00 0x00 0x00 0xffe00000 0x00 0x10000 0x00 0x20000 0x00 0xffe20000 0x00 0x10000 0x01 0x00 0x00 0xffe90000 0x00 0x10000 0x01 0x20000 0x00 0xffeb0000 0x00 0x10000>;

	r5f@0 {
		compatible = "xlnx,zynqmp-r5f";
		reg-names = "atcm0\0btcm0";
		reg = <0x00 0x00 0x00 0x10000 0x00 0x20000 0x00 0x10000>;
		memory-region = <0x168 0x16c 0x16a 0x16b>;
		power-domains = <0x81 0x18110005 0x81 0x1831800b 0x81 0x1831800c>;
		mboxes = <0x16e 0x00 0x16e 0x01>;
		mbox-names = "tx\0rx";
	};

	r5f@1 {
		compatible = "xlnx,zynqmp-r5f";
		reg-names = "atcm0\0btcm0";
		reg = <0x01 0x00 0x00 0x10000 0x01 0x20000 0x00 0x10000>;
		memory-region = <0x169 0x171 0x16f 0x170>;
		power-domains = <0x81 0x18110006 0x81 0x1831800d 0x81 0x1831800e>;
		mboxes = <0x172 0x00 0x172 0x01>;
		mbox-names = "tx\0rx";
	};
};

openamp_mailbox@ff360000 {
	#address-cells = <0x02>;
	#size-cells = <0x02>;
	compatible = "xlnx,versal-ipi-mailbox";
	reg = <0x00 0xff360000 0x00 0x20 0x00 0xff3f0a00 0x00 0x1ff>;
	xlnx,ipi-id = <0x05>;
	interrupts = <0x00 0x21 0x04>;
	reg-names = "ctrl\0msg";
	ranges;
	interrupt-parent = <0x4b>;
	phandle = <0x16d>;

	mailbox@ff340000 {
		phandle = <0x16e>;
		compatible = "xlnx,versal-ipi-dest-mailbox";
		#mbox-cells = <0x01>;
		reg-names = "ctrl\0msg";
		reg = <0x00 0xff340000 0x00 0x20 0x00 0xff3f0600 0x00 0x1ff>;
		xlnx,ipi-id = <0x03>;
	};

	mailbox@ff350000 {
		phandle = <0x172>;
		compatible = "xlnx,versal-ipi-dest-mailbox";
		#mbox-cells = <0x01>;
		reg-names = "ctrl\0msg";
		reg = <0x00 0xff350000 0x00 0x20 0x00 0xff3f0800 0x00 0x1ff>;
		xlnx,ipi-id = <0x04>;
	};
};`

[tnmysh]

@yybingyybing

If RPMsg channel is not created, That probably means IPI interrupt is not generated on linux side.
You can do: cat /proc/interrupts on linux terminal, and see interrupt list. Check for openamp_mailbox interrupt number. Also, for vek280 instead of compatible = "xlnx,zynqmp-r5f"; you should use versal compatible. (Please check bindings).

If you are using petalinux and generating all the artifacts using petalinux, then everything should work out-of-the box. I don't understand why it's not working for you.

Can you attach full system.dtb ?

[tnmysh]

Can you attach full system.dtb ?

@yybingyybing please ignore above request. I don't need full system.dtb. Let me know if you are able to find out the problem on why interrupt is not generated on APU side. Most probably the issue is not correct IPI mask is used on RPU side:

https://github.com/Xilinx/embeddedsw/blob/9a17886209aaf5c0a5f50f58b6cf4fe298911159/lib/sw_apps/openamp_echo_test/src/machine/zynqmp_r5/platform_info.h#L50