Skip to content
This repository has been archived by the owner on Nov 1, 2023. It is now read-only.

2021.01.11: Working Group Meeting

Jump to bottom
Al Stone edited this page Jan 14, 2021 · 1 revision

We didn't really have a formal agenda this time. We decided to discuss these three topics:

  • Current state of the hypervisor work
  • PMU addition to the SBI spec
  • Next steps:
    • Goal: produce initial draft of Linux spec
    • Questions to be investigated:
      • ISA: RVA20? Is that a sufficient description?
      • EBBR: can this be re-used wholesale?
      • Minimal hardware: what is the least common denominator for hardware to run Linux?

That's where we started the discussion. However, all the interest was in the goals for some concrete work on the spec.

The first conclusion reached was that we need to define two platforms; these are the highest priority:

  1. a Linux Development platform -- this would be a platform where we define a reasonable set of things that balances between what people are shipping today, and what those really should be longer term. The idea would be to essentially define a "Level 0" and then create additional, possibly more restrictive levels as more of the hardware standards develop. For example, we have the PLIC/CLINT today, but those may be replaced very soon -- so perhaps the spec requires the new interrupt controller instead of PLIC/CLINT. This would require the A-profile (probably RVA20, but plan for RVA22).

  2. a Microcontroller platform -- think, "embedded systems." This would require the M-profile (again, RVM20 likely, but may need to start with RVM22 just due to timing).

The second conclusion was that there is a basic set of requirements needed for each platform:

  • ISA -- which profile, with what constraints?
  • BIs -- which binary interfaces are required?
  • Processor discovery -- how do we determine what flavor of the RISC-V ISA is being used, and what parts have or have not been implemented?
  • Hardware discovery -- how do we determine the hardware available to the entire system? Device Tree (aka, Open Firmware) and ACPI are the primary technologies in use today.

This led to some discussion on the differences between processor configuration discovery being worked on by the config group, versus SoC (or board, or system) configuration discovery. That in turn led to a discussion of what to do with the processor config and how to discover it, but we basically ran out of time, at that point.

The actions that came out of this are that a small group started working on specifying ISA constraints off-line (we'll move it to the mailing list ASAP), and that your author will start on a brief outline; hopefully, that will provide a starting point where others can come in and flesh out the content.

Clone this wiki locally