Re: [PATCH v15 22/23] x86/mce: Improve error log of kernel space TDX #MC due to erratum

[Dave Hansen]

On 12/5/23 12:41, Borislav Petkov wrote:
> On Tue, Dec 05, 2023 at 08:33:14PM +0000, Huang, Kai wrote:
>> Yes I understand what you said.  My point is X86_FEATURE_TDX doesn't suit
>> because when it is set, the kernel actually hasn't done any enabling work yet
>> thus TDX is not available albeit the X86_FEATURE_TDX is set.
> You define a X86_FEATURE flag. You set it *when* TDX is available and
> enabled. Then you query that flag. This is how synthetic flags work.
> 
> In your patchset, when do you know that TDX is enabled? Point me to the
> code place pls.

TDX can be "ready" in a couple of different ways:

1. The module is there and running SEAMCALLS (tdx_platform_enabled())
2. The module is initialized and ready to run guests.  This happens
   after init_tdmrs() and init_tdx_module() return success.

#1 is known at boot.
#2 doesn't happen until just before KVM runs the first TDX guest.
Here's the patch for #2:

> https://lore.kernel.org/all/566ff8b05090c935d980d5ace3389d31c7cce7df.1699527082.git.kai.huang@xxxxxxxxx/