Open Source Summit North America 2025

The test harness (aka board farm) was initially designed to enable automated testing of Linux drivers on actual hardware. This ultimately became a way to implement continuous testing on hardware, every commit or PR, a build and test is run, and results are reflected back to the commit. It is designed to cater a wide range of peripheral devices and platforms mainly for Analog Devices' Kuiper Linux distribution which is built with more than 1000 Linux device drivers compatible with Xilinx and Intel FPGAs, Raspberry Pi boards, and several other platforms.

This talk covers the design and implementation of such a fully automated test harness. The implemented architecture leverages the use of readily available components/technologies such as Jenkins, Docker, NetBox, and JFrog Artifactory and, at the same time, includes custom-built tools that can be tailored and extended to support existing or new devices and platform types.

This talk will also cover how the team handled the challenges encountered while implementing and addressing features such as resource queueing to minimize waiting time of running builds and also how we make use of sdcardmux to handle recovery.

Virtio-msg is a new virtio transport that allows Virtio to be used on AMP systems between Linux and Zephyr on a co-processor, between the Linux Kernel and Secure World TEEs like Trusty and OP-TESS, and other places where Virtio-mmio and Virtio-pci do not work.

Media Request API was designed to allow V4L2 devices like memory to memory video processing or codec devices and complex camera pipelines to support per frame settings and per frame metadata. In this talk we will deep dive on the details of this API with example use cases such as memory to memory video processing devices and complex ISP pipelines using this API for per frame settings and per frame metadata.

The Yocto project offers unparalleled flexbility in how you structure your embedded Linux image builds. Yocto's multilayered approach enables us to build different images for different purposes (e.g. development, production, etc). This flexibility extends beyond the image build process out to runtime.

What if you could load debug symbols onto an edge device? Restart your application with gdbserver? Start bpftrace sessions? Connect to a VPN? All at the press of a button? You can through the magic of systemd and overlayfs. systemd-sysext activates and deactivates system extension images merging them together to compose the contents of your /usr/ and /opt/ directories.

By separating your root filesystem into individual layers that each do one thing and one thing only, you can respond to trouble by reconfiguring your system at runtime. I will show you how to leverage system extension images and btrfs subvolumes to build Yocto images that can be reassembled on demand.

OP-TEE is an open source Trusted Execution Environment (TEE) designed as a companion to the rich Linux operating system environment. Memory safety is critical to OP-TEE because its applications provide secure interfaces that can be used by the (potentially compromised) rich OS to manipulate user's/system credentials, keys and confidential data. In particular Rust's memory-safety guarantees can be a huge step forward to harden the OP-TEE environment against any memory based exploits.

This session will focus on the evolution of Rust within the OP-TEE ecosystem, the story to make Rust a first class citizen for writing OP-TEE Trusted Applications (TAs) and how we can build Rust TAs at performance parity with their C counterparts. As well as reviewing the work to date, we'll also look at some options for leveraging Rust's memory- and thread-safety features within OP-TEE core itself: should we start with enabling pseudo trusted applications to be written in Rust?