This repository has been archived by the owner on Nov 1, 2023. It is now read-only.
-
Notifications
You must be signed in to change notification settings - Fork 39
2020.10.19: Working Group Meeting
Al Stone edited this page Oct 22, 2020
·
1 revision
Details: https://sites.google.com/a/riscv.org/risc-v-staff/home/tech-groups-cal
- Proposed scope/charter (revisit)
- https://github.com/riscv/riscv-platform-specs/wiki/Draft-Charter
- Being submitted: just a quick mention that this was happening
- SBI Reset extension (revisit)
- discussion: https://lists.riscv.org/g/tech-unixplatformspec/message/212
- discussion: https://lists.riscv.org/g/tech-unixplatformspec/message/226
- Lengthy discussion on whether or not SBI can be approximately equal to
the existing ARM PSCI spec since both serve similar purposes:
- Pros: no need to re-invent some things, may be able to re-use existing test suites and tools
- Cons: OpenSBI already exists (along with a couple of other implementations) so they would need rework, which begs the question of why replace what is already in use
- Continued discussion: reviewed the prototype re-use of PSCI in Linux to implement SBI and consensus view was that it could work, but did not seem to improve on what we have in the kernel already; deferred a final decision until we could have further discussion with Paul.
- SBI PMU extension (revisit)
- discussion: https://lists.riscv.org/g/tech-unixplatformspec/message/197
- See SBI Reset above
- Proposal: Hart Suspend Extension for IDLE (new)
- discussion: https://lists.riscv.org/g/tech-unixplatformspec/message/274
- This needs a little more discussion on the mailing list, but is really a derivative of work already in progress, so is probably fine; deferred final decision.
- Proposal: Introduction to spec (new)
- discussion: https://lists.riscv.org/g/tech-unixplatformspec/message/290
- Talked over the concepts being introduced:
- in particular, the definition of profiles and how they might be used
- noted that some of the elements of the profile n-tuple such as Firmware Interface and Boot Interface need crisper definitions -- e.g., if we use UEFI for the firmware interface, how do we distinguish UEFI+DT vs UEFI+ACPI? Or should Device Enumeration Interface be separate? For the boot interface, do we need to be more specific, as in UEFI Shell vs UEFI, or bootm and booti vs U-Boot? Al will draft v2 and try to address these.
- there will be a need to version everything, both the profiles and the interfaces they use.
- will need to have short names -- e.g., LinuxServer Profile -- vs having to repeat the n-tuple in detail
- there may be some interfaces, such as SBI, that we may decide are common and must be used by all profiles; those will need to be called out in a section of the spec
- the n-tuple (with versioning) could be passed into the kernel via the
DT Chosen node; this probably belongs in the boot protocol document
- this led to a discussion of where to document the Linux boot protocol and this best fits into the Linux documentation subtree; will need to talk with Palmer about writing this down
- should boot protocol/interface be a separate element in the n-tuple defining the profile? possibly so, but not immediately clear.
- in general, this intro was seen as a good way to define and use profiles, and helps generate the structure for the remainder of the spec
- Proposal: Add RustSBI into 'SBI Implementation IDs' list (new)
- discussion: https://lists.riscv.org/g/tech-unixplatformspec/message/338
- this appears to fall in the "mostly obvious" category, so best to just merge it
- Q: what's happening with the PRs against the SBI doc? A: in the past, they would be discussed on list, then patches made once consensus was reached, then PRs made; at some point, Atish and/or Palmer would merge the PRs after review. Periodically, Palmer would make a release. Atish will look back through the PRs and take care of some of the mostly obvious requests, and do some review; Al will also provide some review. Will need to see about periodic releases and evaluate whether it's time for another soon -- this all sort of depends on how many substantive changes are being made.