On Sent: Tuesday, August 9, 2022 9:14 AM, Borislav Petkov wrote: > On Tue, Aug 09, 2022 at 02:36:33PM +0000, Kani, Toshi wrote: > > The logic needs to be latched on ghes presence, i.e., the condition is > > If GHES is not enabled, there'll of course be a stub which returns > false. Right. Just checking since the example pseudo code did not have it. > > Agree that changing all edac drivers to check with GHES is an option. > > Not all - all relevant drivers for your - HPE - use case. We don't load > ghes_edac on anything else. Known-good platforms only, remember? I think this should be all edac drivers on x86. > > In this approach, though, they will need to check with foo_preferred() > > when a new FW interface FOO is introduced. > > I'm afraid I don't understand what you mean here. I was referring a hypothetical future case that ACPI GHES might not be the only FW interface for FF-based memory error reporting table going forward. Toshi
