Tock OT Notes 2/21/2020

Present:

• Laura Abbott, Oxide Computer
• Brad Campbell (Chair), University of Virginia
• Jon Flatley, Google
• Alistair Francis, Western Digital
• Garret Kelly, Google
• Patrick Mooney, Oxide Computer
• Bryan Cantrill, Oxide Computer
• Adam Leventhal, Oxide Computer

TODO

• @gkelly: Open PR to add Silvestrs to the WG
• @gkelly: Investigate options for LLVM development
• @bradjc: Add libtock-rs RISC-V support to charter
• Add @jrvanwhy to the call
• all: Work to streamline libtock-rs building for RISC-V platforms

PR Merge Policy

• Brad: tock-core has two types of PRs: upkeep and significant. Upkeep are merged quickly with one or two approvals, significant are merged after all team members approve or after at least a weak with no objections. Anyone can block a PR and then further review and discussion occurs before merging.
• Consensus this is a reasonable policy to mirror for OT WG.
• Makes sense to use the same approach.

libtock-rs

• RISC-V support not well maintained
  • Scripts not correct
  • Very difficult to build PMP test
  • Alistair: works if you build manually
• No CI support currently
  • Something we should work towards
  • Can use QEMU to help
• Goal: converge on a build system to make building libtock-rs easy on RISC-V to help ensure it gets more testing as we work towards CI
• Google: expect libtock-c to be supported on OT, but want top notch support for libtock-rs

libtock-rs ownership

• Not clear that libtock-rs should be completely under the OT WG umbrella
• However, there is a lot of overlap between OT WG and libtock-rs development
  • Decided to see if we could add RISC-V support in libtock-rs under the OT WG umbrella in the same way that RISC-V support for the kernel is.

Relocation Support

• Bryan: not having relocation for RISC-V/libtock-rs is an "iceberg" standing in the way
• Alistair: Difficult to have to change how apps compile when kernel size changes
  • Maybe could automate the process with some tooling?
• Alistair: Western Digital is working on GCC PIC support for RISC-V
  • Currently in the design phase
  • Working on an ABI
• Garret: Need a better scope of work for what updates LLVM needs
  • Can look for person/team at Google who could help
• Brad: How feasible is updating LLVM to do the correct relocations?
  • Laura: It seems like it is doable by an experience LLVM engineer
• Garret: Google intends to use monolithic signed binaries initially to workaround the relocation problem

Tock 2.0

• Would include breaking changes to the syscall interface
• Brad: Tock 1.0 specified the low level syscall interface, the TBF format, and a small handful of driver interfaces
• Brad: consensus is to do a 2.0 version which would aggregate several breaking changes to the syscall interface
• Brad: timeline:
  • Now/next few weeks: suggest ideas for syscall changes
  • 1+ months: discussion of proposed changes, work towards a design
  • 3-6 months: implement and test changes
• Brad: loading apps at runtime should not require 2.0