ContextQMD
Back to Tock

Tock OT Notes 4/9/2020

doc/wg/opentitan/notes/2020-04-09.md

latest2.5 KB
Original Source

Tock OT Notes 4/9/2020

Present:

  • Brad Campbell (Chair), University of Virginia
  • Phil Levis, Stanford
  • Silvestrs Timofejevs, lowRISC
  • Johnathan Van Why, Google
  • Alistair Francis, Western Digital

Updates

  • Silvestrs: Working on a PR for conditional compilation. OT has two targets which differ very little, and this would allow the same board to compile for them.
  • I started with a simple approach, now looking for feedback.
  • Brad: One question is how much overhead is worth it when hardware differs in small ways. We don't currently have an approach for this, so your PR is pretty experimental. We've done the nested crates approach with the nRF, but that seems heavyweight for this application.
  • Another question is should the config options be relatively fixed, or should it be easy for users to provide their own configuration settings?
  • Silvestrs: For debugging, may be useful to set a custom frequency.
  • Brad: I always worry about cfg options which can lead to a very complex code base, but here the values should at least be very clear since they are based in hardware. It is useful to have a default setting so that cargo tools will work even if there are no cfg options set.
  • Silvestrs: Config files might be better for providing more structure.
  • I'm still experimenting with Makefile-provided cfg features that get passed down.
  • Brad: I agree on config files, but I think how things get set is still tricky because having too many different configuration methods can get confusing.
  • Brad: Why the configurable baud rate?
  • Silvestrs: I wasn't sure either, but it seems the developers use different baud rates on different platforms to get better performance on the FPGA.

Oxide

  • Phil: Oxide is stepping back from active Tock development.
  • After learning more about Tock and evaluating their use cases, they don't see Tock's design as a good fit for their needs. And they don't want to try contort the project to their intended use case.
  • Bryan did indicate that they do like the design of the kernel, and intend to still use it in some form.
  • But the overhead and complexity of the syscall interface for libtock-rs is not needed in their design. They want to run everything in Rust, and want blocking operations closer to the kernel.

DIFs

  • Silvestrs: I have proof of concept, "hello world" style C DIF integration with Tock working.
  • Brad: Excellent, I know that is an important goal.

HMAC