Written by Igor Stoppa, Senior Software Architect at Nvidia
For more than two decades, Linux has made inroad in new fields of applications, from data centres, to embedded. We see now a growing demand for Linux in safety critical applications, ranging from automotive to robotics, to medical appliances.
However, Linux was not designed with these applications in mind, and unsurprisingly it is not an ideal fit, at the moment.In particular, one major pain point is the very limited resilience to spatial interferences originating from within the kernel itself.
Furthermore, the code base if much larger than what can be found in other operating systems traditionally found in safe applications. This is also compounded by the fact that Linux does not follow the processes traditionally in use for Functional Safety.
Summary
In the video, I describe my ongoing experiment of modifying the Linux kernel, to introduce a form of Address Space Isolation, meant to provide a mechanism enforcing freedom from interference. The presentation describes the problems, possible means to address it, and the current progress with the implementation. You’ll see a methodology for the safety analysis of a Linux system and mechanism for improving the safety of selected components.
This presentation ties both into the scope of the Linux Features for Safety-Critical Systems Working Group and the Critical SW track at Open Source Summit Europe. Though this work is not formally sponsored nor endorsed by ELISA, it is something I shared with the community for brainstorm and discussion purposes.
If you’d like to learn more about the Linux Features for Safety-Critical Systems Working Group or you’d like to continue this conversation, please join the mailing list or a WG meeting here.