This commit supports spdm-requester-emu communicate with other endpoints

[NguyenTanNhutQuang]

via MCTP in kernel. It also supports sending single command GET_VERSION to test the interface.

Tested:
Send command:
spdm_requester_emu --trans MCTP_KERNEL --eid 20 --cmd GET_VERSION
Sample output:
spdm_requester_emu version 0.1
trans - 0x5
spdm_cmd - 0x01
connect success!
context_size - 0x21d8
libspdm_send_spdm_request[0] msg SPDM_GET_VERSION(0x84), size (0x4):
0000: 10 84 00 00
Platform port Transmit transport_type: 00 00 00 05
Platform port Transmit size: 00 00 00 05
Platform port Transmit buffer:
05 10 84 00 00
Platform port Receive buffer:
05 10 04 00 00 00 03 00 10 00 11 00 12
libspdm_receive_spdm_response[0] msg SPDM_VERSION(0x4), size (0xc):
0000: 10 04 00 00 00 03 00 10 00 11 00 12
Client stopped

[jyao1]

@NguyenTanNhutQuang , please fix CI failure.

Please note that spdm-emu need to be built in both Linux and Windows OS environment.
I feel you need to have a way (maybe MACO) to exclude Linux kernel code if it is Windows/Visual Studio.

[NguyenTanNhutQuang]

@NguyenTanNhutQuang , please fix CI failure.

Please note that spdm-emu need to be built in both Linux and Windows OS environment. I feel you need to have a way (maybe MACO) to exclude Linux kernel code if it is Windows/Visual Studio.

Hi jyao1,

Is there anyway that I can check CI on my local server before push to github? Thank you

[jyao1]

@steven-bellock , do you have any idea?

[steven-bellock]

In terms of bandwidth / utilization it's fine to test your changes by pushing to this repository. The CI/CD utilization of this repository is low. If you really want to run locally you'd need something like https://github.com/nektos/act.

[mcsmonk]

Could we consider using the official Linux man page — such as https://www.man7.org/linux/man-pages/man7/mctp.7.html — instead of the Discord link?

I’m unable to access Discord within the company network, and the man page provides a more reliable, standardized, and accessible reference for technical documentation.

[mcsmonk]

To clearly distinguish the Linux kernel’s MCTP stack, would it be better to rename SOCKET_TRANSPORT_TYPE_MCTP_KERNEL to SOCKET_TRANSPORT_TYPE_MCTP_LINUX_KERNEL?

Additionally, since other transport layers may be added in the future, it would be advisable to use platform-specific qualifiers in all MCTP transport type definitions. For example:

Suggested change

	#define SOCKET_TRANSPORT_TYPE_MCTP_KERNEL 0x05
	#define SOCKET_TRANSPORT_TYPE_MCTP_LINUX_KERNEL 0x0101 // MCTP transport via Linux kernel stack,

This approach ensures clarity, and supports scalable extension for new transport backends.

[wmaroneAMD]

I have a branch here that is derived from this:

https://github.com/wmaroneAMD/spdm-emu/tree/ntnq-kernel

It has been rebased and addresses the build comments here. I haven't tested the windows build just yet, but I'd like to move this feature forward as it is immediately useful to me.

@mcsmonk Your last comment I have not yet integrated but intend to do so.