| RVI_Disclosures_Female_Short.mp4 |
|
List the agenda items or include them in the other materials or presentations being used during the meeting.
current version: https://lf-riscv.atlassian.net/wiki/download/attachments/217448450/fusa-whitepaper.20250204.draft.pdf?api=v2
current version with proposed introduction: https://lf-riscv.atlassian.net/wiki/download/attachments/217448450/fusa-whitepaper.20250204.introduction-draft.pdf?api=v2
whitepaper GitHub repository: https://github.com/riscv/sig-functional-safety-whitepaper
shared drive with all the input chapters: https://drive.google.com/drive/folders/13cxh4IxDGbEsit0rs7yuihduq85MrvOX?usp=drive_link
Disclosures
Update on the whitepaper status
Nothing new
Waiting for Partitioning chapter
Request participants contribution
Community activities that can have impact on the SIG activities
Performance events TG: Proposing a default encoding for a minimal subset of performance events.
Hypervisor SIG: ACPI vs ISA WARL discovering mechanism might disagree.
Radim: the ACPI and ISA should align. It’s an issue for virtual machines migration, where the software might rely on WARL and not the ACPI, which might break the migration which should follow the ACPI. Shouldn’t be an issue with safety critical systems, where migration is not needed.
Daniel, question: could be it an issue with the multithreading control gap?
Radim: It shouldn’t. The software would have to ask the machine mode to activate the harts, and the ACPI and the machine mode code should be aligned, i.e. or both indicate that the multithreading support is enabled or disabled, but not one indicating that it is enabled and the other one indicating that it is disabled.
Paul, question: to which kind of multithreading are we referring
Daniel: to a core shared by two different harts, not to harts running in different cores.
Holger: in safety systems they have worked on the multithreading could be disabled through BIOS, no need to control the multithreading through the ISA.
Profiles extension: The Ssstrict extension is defined in the RVA23/RVB23 profiles (as profiles extensions).
Ken: It might be extended to the regular ISA (i.e. outside of the profiles).
Radim/Ken: it will be detectable by software (device tree, ACPI, etc.), but not something that can be set or unset through CSRs.
Ken: should be mandatory in the automotive profile.
Gaps table discussion
Reminder of previous meeting discussion: reduced watchdog gap to low priority and confirmed the priorities of the redundancy (low priority).
Holger: we can remove the multithread control gap as we have discussed that could be addressed through solutions like ACPI.
Remove the watchdogs gap.
Thomas, question: if we remove the watchdogs gap, does it mean that there are no gaps in the partitioning interrupts topic the gap was identified?
Daniel: No. New gaps can be added, even for topics that were removed from the gaps table, like the partitioning interrupts.
Thomas: Partitioning interrupts and/or ensuring freedom of interference for interrupts are issues that are very important for functional safety. For the next session, Thomas will take a look and see if something is missing and needs to be addressed as a gap.
No news on the reset of the temporal state (fence.t TG). Next fence.t TG meeting on 11 march.
Holger/Bing: Not clear how the material is being shared
Daniel: will add a link to the Confluence page of the Functional Safety SIG and send an email on the means the materials are shared within the SIG.
Holger: in the EU project TRISTAN a partner has proposed a reduced subset of RERI for embedded subsystems. Holger proposes them to present us their work.
Paul: not sure the watchdog standardization gap should be removed.
Thomas: not really. Is very much implementation dependent and the use cases very broad. The way it' s used it can be managed through SBI if necessary.
Meeting recording and transcript here
Title | Presenter | File | |
|---|---|---|---|
| 1 | Meeting support slides | Daniel |
