Tock Core Notes 2023-08-18

Attendees

• Amit Levy
• Brad Campbell
• Alyssa Haroldson
• Alistair Francis
• Chris Frantz
• Leon Schuermann
• Pat Pannuto
• Johnathan Van Why
• Hudson Ayers

Updates

• Amit: Ethernet over USB implemented and working on nRF52.
• Brad: Expanding clippy lint allows. Clarify which clippy lints we need to allow for clippy to pass.
  • Will be PRs addressing some. Feedback will be helpful.
  • Chris: Which group is not a good fit?
  • "restriction"
• Pat: Tyler opened PR for software acks for nRF52. Next up is support for Tock as sleepy end device for Thread.
  • Hudson: will send imix, but

PMP Redesign

• https://github.com/tock/tock/pull/3597
• Two implementations: PMP and ePMP (OT). Some issues and duplications.
• Goal: separate implementation for pmp logic from specific impl of tock process regions.
  • Simple PMP implementation
  • OT/earlgrey implementation
• Alyssa: should we get a review from Vadim? A: yes.
• Current implementation not matching hardware as riscv has evolved. Boot stages can affect how the PMP is configured.
• Downside
  • No longer a single implementation that can be used on any chip.
  • On complex chips, have to implement bottom half of ePMP yourself.
  • ePMP moved to chip folder, must be duplicated for each custom chip.
    • But we currently only have 1 (OT) which is very unique
    • There may be a generic ePMP implementation for future chips. ePMP is standard.
    • earlgrey version of ePMP: board or chip? all earlgrey will use the same bootrom that sets up regions.
• New modular design sets us up to support ePMP in the future.
• Amit: need to move low level conversation to issue or separate discussion.
• Brad: let's implement ePMP when we have an ePMP chip. This seems like good progress.
• OT essentially customizes the ePMP by setting up locked regions.
  • How much does this configuration complicate the implementation?
  • Difficult to do and ensure correctness.
• Open tracking issue.
• Move to OT call.

Maintaining 3.0 Changes

• https://github.com/tock/tock/pull/3622
• What is our version policy?
• How do we maintain changes for major version changes?
• Action: read policy.

Command 0

• https://github.com/tock/tock/issues/3375
• TRD104 specifies that command 0 returns Success if driver exists. A couple stabilized drivers return Success(u32) with the number of available resources.
• What to do?
  • Change to match TRD104: breaking change.
    • Makes the simple check clear and easy to do.
    • Can implement the check in the kernel.
  • Update TRD104.
    • Makes sense to get the same functionality plus more.
    • Seems more elegant.
• Advantage to changing TRD104: no code needs to change. Stabilized drivers stay the same.
• Brad: what is the motivation to change TRD104? Are there new use cases?
• TRD104 written after drivers.
• Command 0 is a safe check to determine if a driver is present.