On 07.03.2025 09:48, Jan Beulich wrote:
> On 07.03.2025 00:35, Andrew Cooper wrote:
>> We've already scanned features by the time init_e820() is called. Remove the
>> cpuid() calls.
>>
>> Signed-off-by: Andrew Cooper <andrew.coop...@citrix.com>
>
> Reviewed-by: Jan Beulich <jbeul...@suse.com>
>
>> Backporting. Not sure it's worth backporing, but it is safe (just) to
>> backport past commit 365f408339d3 ("x86/boot: Load microcode much earlier on
>> boot"). That commit was the last one to reposition early_cpu_init().
>
> At least I wouldn't consider such cleanup to be an obvious backporting
> candidate.
>
>> I'm pretty sure that all 64bit CPUs have MTRR, but I'm less certain if
>> dropping the check is wise given the variety of VM configurations that exist.
>
> We did consider exposing PAT-only configurations to guests, so I don't think
> we should be implying MTRR from being 64-bit (unless we know
> cpu_has_hypervisor
> is false).
Except that - we do:
#define cpu_has_mtrr 1
I guess that wants undoing (pre-dating the consideration of Xen running
virtualized itself).
Jan
