Re: [PATCH v3] x86/mce: Do not treat cache hierarchy errors as memory errors

[Sohil Mehta]

On 10/19/2026 2:06 PM, Luck, Tony wrote:
>> Looking at the comment on top of tdx_dump_mce_info() and commit
>> 70060463cb2b ("x86/mce: Differentiate real hardware #MCs from TDX
>> erratum ones") could poison consumption be reported with a cache
>> hierarchy code?
> 
> Yes. All recoverable poison consumption error codes are "Cache Hierarchy Error"s.
> with signature 000F 0001 RRRR TTLL. It's an uncorrected error, so the recovery
> code will offline the page.
> 
> BUT ...
> 
> Each should be accompanied by a "Memory Error" with signature 000F 0000 1MMM CCCC
> in a different machine check bank (this report originates in the memory controller that detected the ECC issue).
> 

Okay makes sense. I am fine with the change in this patch now.

Reviewed-by: Sohil Mehta <sohil.mehta@xxxxxxxxx>

If needed, the TDX side can change tdx_dump_mce_info() in case the
desired "TDX private memory error. Possible kernel bug." doesn't get
displayed.

> This other report should be used to track uncorrected memory errors per-DIMM. Per-channel etc.
> 



> =Tony