Re: [RFC EDAC/GHES] edac: lock module owner to avoid error report conflicts

[Mauro Carvalho Chehab <mchehab@redhat.com>]

Em Fri, 2 Nov 2012 00:54:57 +0100
Borislav Petkov <bp@alien8.de> escreveu:

> On Thu, Nov 01, 2012 at 11:47:52PM +0000, Luck, Tony wrote:
> > > Right, but at least in the csrow case, we still can compute back the
> > > csrow even with the interleaving, after we know how it is done exactly
> > > (on which address bits, etc). I think this should be doable on Intel
> > > controllers too but I don't know.
> > 
> > No. Architecturally all Intel provides is the physical address in MCi_ADDR.
> > To do anything with that you are into per-system space, and the
> > registers that define the mappings are not necessarily available
> > to OS code ... sometimes they are, and sometimes they are even
> > documented in places where Mauro can use them to write an
> > EDAC driver ... but there are no guarantees.
> 
> One more reason that we need some sort of tables telling us which
> rank/csrow maps to which DIMM and thus silkscreen label so that we can
> be able to say the following from software:

If you take a look at sb_edac driver, you'll see that most of the driver's
logic is the part that do address->DIMM location mapping. It is a very
complex logic.

> 
> "You just had a single corrected ECC error in the DIMM with label
> P0_DIMM_A"

Converting from a DIMM location into a DIMM label is trivial, through:
just a simple table lookup.

As APEI/GHES seem to provide the DIMM location, a simple table lookup
should also work there, as I said before.
 
> or whatever unique naming each platform vendor comes up with.
> 
> The day hw people give me this, I'm going to throw a big party and
> invite all LKML.
> 

-- 
Regards,
Mauro