Summary

LDCG is now holding a vote on prioritisation for the Server Architecture team for the next cycle BUD20-SJC20. 

Please indicate your organisations priority by selecting either option 1 or 2 by Reply All email before the close of business on the 11th of March so we can discuss the results at the LDCG-SC meeting on the 12th of March.


Option 1)
Continue with work on SBSA QEMU and progressing to start of RAS
support for SBSA QEMU machine.

SBSA QEMU Machine is an attempt to create a server level (SBSA)
machine in QEMU as no machine of that class existed before. The
current work is implemented upto approx SBSA level three (ARM 8.0
architecture) although we have not quite got to the point we can run
the ACS on it yet. There is work in the backlog to extend this to
later levels if there is resource available in the future.

There are two work items in the next cycle.

https://projects.linaro.org/browse/LDCG-777

Investigate and then implement if upstream is open to the idea a
method to trigger hardware errors in the QEMU machine that can then be
picked up by the secure firmware implementation and passed through the
firmware stack to the kernel in the same method as would happen on
real hardware. Giving the ability to test the full stack for such
events which is currently dificult without expensive test hardware.

https://projects.linaro.org/browse/LDCG-778

Improve the ACPI support in the QEMU machine to dynamically generate
tables so hardware can be dynamically added/removed from the machine
via the QEMU cmdline/config files rather than a hardcoded set of
hardware. This would take QEMU internal description of the hardware
and use the recently upstreamed Dynamic ACPI support to generate the
tables.

Option 2)
Intel implemented the 0day system for their CPUs which carried out
various testing on patches submitted to LKML (and subsystem specific
lists). Loosely described these tests covered Patch quality (does it
pass kernel coding standards), static analysis (looking for common
errors and security issues), regression testing (does the patch break
the kernel), and performance regression testing (does the patch
greatly affect performance for certain workloads). This work from
Intel was never open sourced and is not available for the community to
extend or improve.

Linaro is the ideal place to do such testing and test development for
the ARM community. Currently Linaro is running the LKFT testing which
is being extended to servers right now inside the LDCG COLO.

There is future work to scope out the sort of tests we would want to
implement for an open source 0day for ARM. And then to continue to
implementation of these tests. And building the infrastructure
required to run these tests (testing farm etc).

https://projects.linaro.org/browse/LDCG-890
https://projects.linaro.org/browse/LDCG-891 

This vote was taken via email sent by Graeme Gregory to ldcg-sc-members on 3 March 2020. 

Results