Re: [PATCH] x86/mce: Add workaround for SKX/CLX/CPX spurious machine checks
From: Borislav Petkov
Date: Wed Feb 16 2022 - 13:53:06 EST
On Wed, Feb 16, 2022 at 06:41:58PM +0000, Luck, Tony wrote:
> > Well, we could try to decode the instructions around rIP when the #MC
> > is raised and see what caused the MCE and perhaps pick apart which insn
> > caused it, is it accessing behind the buffer boundaries, etc.
>
> Is this a case of "perfect is the enemy of good enough"?
Well, you guys sounded like this happens left and right...
> It is a rare scenario (only a pain point for Jue because Google has
> billions and billions of cores running this code). You need:
>
> 1) An uncorrected error
> 2) That error must be in first cache line of a page
> 3) Kernel must execute page_copy from the page immediately before that page
>
> When all three happen, kernel crashes because we don't
> have a recover path from kernel page_copy
You should've lead with that - this is basically one of those "under a
complex set of conditions" things.
Anything against me adding them to the commit message?
--
Regards/Gruss,
Boris.
https://people.kernel.org/tglx/notes-about-netiquette