On 18.01.2022 21:28, Andrew Cooper wrote: > On 17/01/2022 15:30, Jan Beulich wrote: >> As of SDM revision 076 there is a CPUID bit for this functionality. Use >> it to amend the existing model-based logic. >> >> Signed-off-by: Jan Beulich <[email protected]> > > https://lore.kernel.org/lkml/[email protected]/T/#u > suggests that Sapphire Rapids also needs the model specific treatment.
Well, I can go and pull in their a331f5fdd36 just to be on the safe side. I have to admit that it's not clear to me whether that older commit is made obsolete by the CPUID bit check about to be added. > I agree with the "only-expose-on-error" observation, so perhaps we ought > to make these details available to the hardware domain in a suitable form. hypfs? Jan
