Re: [PATCH 2/8] x86/boot: Collect AMD speculative features earlier during boot

[Andrew Cooper <Andrew.Cooper3@xxxxxxxxxx>]

On 26/01/2022 13:37, Andrew Cooper wrote:
> On 26/01/2022 12:44, Roger Pau Monné wrote:
>> On Wed, Jan 26, 2022 at 08:44:46AM +0000, Andrew Cooper wrote:
>>> All AMD IBRS-related features are in CPUID.0x80000008.ebx.  Collect them in
>>> early_cpu_init() so init_speculative_mitigations() can use them.
>>>
>>> Rework the existing logic structure to fill in c->extended_cpuid_level and
>>> separate out the ambiguous use of ebx in an otherwise 0x80000008-specific
>>> logic block.
>>>
>>> Signed-off-by: Andrew Cooper <andrew.cooper3@xxxxxxxxxx>
>> Reviewed-by: Roger Pau Monné <roger.pau@xxxxxxxxxx>
>>
>> It would be good to update the comment ahead of early_cpu_init to
>> mention it now also gather speculation-related fields from CPUID in
>> order to do early setup of mitigations.
>>
>> I think you could also use boot_cpu_data in spec_ctrl.c print_details
>> instead of fetching again the cpuid leafs?
> Hmm - I may have a mistake here.
>
> Boot time CPUID handling is giant mess, and I haven't had time to finish
> my work to make BSP microcode loading dependent on xmalloc(), allowing

Sorry.  I mean "independent" here.

~Andrew

> it to move far earlier, and removing the early/late CPUID split.
>
> However, init_speculative_mitigations() is called after late CPUID
> setup, so e8b should be suitably collected.  Let me try to figure out
> what's going on.
>
> For print_details(), I have a feeling that may have been an artefact of
> an early version of the logic, and likely can be cleaned up.
>
> ~Andrew