Why there are lots of patches in nvidis sdk kernel?

Robotics & Edge Computing Jetson & Embedded Systems Jetson AGX Orin
1

Hi Nvidia experts,

Recently, I am working a BSP project based on Nvidia AGX orin platform.
Like some BSP development that I worked before, I need to get sdk kernel patches(most of them are related with drivers and arch) to my own kernel. When I started to study Nvidia sdk, I found out there are more that 37000 sdk patches in GitHub - OE4T/linux-jammy-nvidia-tegra, This quantity is astonishing， and I don’t have the courage to port then into my 6.12 kernel. Because from 5.15 to 6.12, there are so large difference. But When I read linux related user guide, I found out below link
Bring Your Own Kernel — NVIDIA Jetson Linux Developer Guide 1 documentation
it says that I can apply these recommended patches into my own kernel, and build it for running on Nvidia platform. Because my kernel is v6.12, I don’t need to apply any patch. So I have a question as below:

  1. If I use the upstream kernel directly, are the Nvidia advanced features, for example, CUDA, Deep Learning and so on, able to be supported well by the upstream kernel ?
  2. why there are more than 37000 sdk patches? Are they the essential ones that I must port them into my own kernel, so that I can support Nvidia advanced feature well?

thanks,
Limeng

3

Just to clarify that the yocto github you are referring to which has 37000 patches are not directly maintained by NVIDIA.

This is a project supported by our partner Ridgerun.

If you want to use upstream kernel, then the patch list is mentioned as our document.

4

ok, got it.
Back to my 1st question, if I use the upstream kernel with all the patches listed in Bring Your Own Kernel — NVIDIA Jetson Linux Developer Guide 1 documentation directly, are the Nvidia advanced features, for example, CUDA, Deep Learning and so on, able to be supported well by the upstream kernel ?

thanks,
Limeng

7

You should know that with L4T R35.x (flashed with JetPack 5.x) and earlier (L4T is just what you would call Ubuntu with NVIDIA content added) the kernel was not mainline and you’d have used that kernel plus out-of-tree content. As of L4T R36.x (flashed with JetPack 6.x) this is the mainline kernel and not NVIDIA’s modified kernel.

8

Hi,

Yes, those sdk shall still work with upstream kernel.

9

thanks for all of your professional answer.

thanks,
Limeng

10

I have another question, is repo GitHub - OE4T/meta-tegra: BSP layer for NVIDIA Jetson platforms, based on L4T maintained by Nvidia?

thanks,
Limeng

11

No, it is from our partner.

12

ok! got it.

thanks,
LImeng

13

This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.