Ecosystem 3.1.11: Patch real-time kernel for linux-5-10-61

[Led]

Hi all

For test purpose I want to run a real-time kernel on the STM32MP157c-DK2 with the newest ecosystem 3-1-11. The goal is to develop a real-time kernel module, out of the Yocto environment.

I did build the Image with Yocto and flashed the image. Apparently the ecosystem 3.1.11 uses the Linux kernel 5-10-61.

I could also successfully build and run the Linux kernel 5-10-61 outside Yocto, adding the .config from the /proc/config of the imag.

Now I'd like to add the reat-time patch, but there is no patch for 5-10-61, but only for 5-10-59 or 5-10-63.

• I tried to use the 5-10-59 patch with the 5-10-61 kernel. It compiled successfully, but the kernel did not start up.
• I also tried to use the newest kernel with the proper patch: 5-10-83. This did not work either.

What do you recommend to do?

By the way: the M4 inside stm32mp1 is intentionally not used.

[Olivier GALLIEN]

Hi @Led​ ,

Sorry but, as already mentioned in post https://community.st.com/s/question/0D53W00000tqn7aSAA/linux-kernel-on-stm32mp157f

ST will not provide any support regarding Linux-RT patches integration.

>> By the way: the M4 inside stm32mp1 is intentionally not used.

Can I ask why, since M4 is - on purpose - present in MP1 to offload real-time constraint task ?

Don't hesitate to elaborate further your need to discuss with you some possible alternative.

Thanks,

Olivier

[Led]

Hi Olivier,

Thanks a lot for the fast reply and the link. I have not seen it before.

May I place my recommendation here that, next time the kernel version is selected, ST also takes into consideration to select a version where a RT-patch is is available, if possible. Just for the people who want to do it on their own, can simply do it.

To answer your question here a bit of background: My current project is not only to implement a real time system, but also to build up knowledge how this can be done without a dedicated processor. In a recent project I planned to use the M4 to read some events and pass them via the internal interface to the Linux application. Finally the effort to implement this interface and supporting different applications running on different processors with different build environments, was too large. In addition the uncertainties about the real-time conditions of this interface was to big. As the project relaxed RT requirements, we implemented everything in standard Linux.

Best regards,

Led