Re: [PATCH] x86/MCE: Set the MCE's status and misc members

["Luck, Tony" <tony.luck@intel.com>]

On Fri, Aug 07, 2020 at 10:05:50AM -0500, Steve Wahl wrote:
> There's one small code change I suggest below.
> 
> But now it's time to think about the description.  We need to zoom out
> to the 10,000 ft. view.  IMHO, to understand what this patch does,
> your current description requires the reader to be more familiar with
> the involved functions than we should expect.

Maybe also travel back in time to why apei_mce_report_mem_error()
was created in the first place.

This was a workaround for a couple of Intel CPUs (Nehalem, Westmere)
that didn't provide a system address in MCi_ADDR. Since BIOS could figure
out the address, it was tasked with reporting to the OS suing APEI.

apei_mce_report_mem_error() did the minimum necessary to create
an error log that mcelog(8) daemon could process.

Jump forward in time to Sandybridge (and all successors) and the
reason for the workaround is gone. All of them provide a system
address in the machine check bank ... so this APEI path is a
poor substitute for getting the data from the machine check bank
(which will give you the channel, the corrected error count, and
perhaps other useful information).

So why are we applying band-aids to this decade old workaround?

-Tony