Tegra4Linux Kernel on TX2


#1

NVIDIA’s Tegra4Linux has a modified kernel to improve performance and make available some TX2-specific features. Does the ResinOS use this kernel or stock Linux on TX2?


#3

Hi @chachi – We use the nvidia kernel, with the latest release being based on the l4t-r28.2 kernel source.

All of our device support work is fully open source, so if you wanted to lift the hood and explore what else is included in our TX2 support the repo is at: https://github.com/resin-os/resin-jetson-tx2 – The kernel used can be seen in https://github.com/resin-os/resin-jetson-tx2/tree/master/layers by following the meta-tegra entry.

The patches we add to that kernel are available at: https://github.com/resin-os/resin-jetson-tx2/tree/master/layers/meta-resin-jetson/recipes-kernel/linux/linux-tegra

Hope that helps to clarify what’s in the TX2 image & let us know if there’s any additional information we can provide


#4

That’s great to hear. Thanks for the pointers to the repos, as well.


#5

Related, but are you aware of any way to easily get the tools (gstreamer, CUDA) which are included in the NVIDIA JetPack installer into Docker containers on top of ResinOS outside of just copying the files from a JetPack installed TX2?


#6

Personally I’ve copied the setup scripting part from Dockerfile.cudabase at https://github.com/open-horizon/cogwerx-jetson-tx2 to get started.

Make sure to change the FROM arm64v8/ubuntu:xenial-20180123 to the resin version, ie FROM resin/%%RESIN_MACHINE_NAME%%-debian, and add your application specific commands and start command at the bottom of your Dockerfile.template.

I haven’t used gstreamer my self so can’t comment on that to much, however https://developer.ridgerun.com/wiki/index.php?title=Compile_gstreamer_on_tegra_X1_and_X2 seems to have some info on how to compile gstreamer with NV support – not sure how relevant it is, but might be worth a try