Re: Patch to read/parse the MPC oem tables

[Randy.Dunlap]

"Martin J. Bligh" wrote:
> 
> >> This patch will parse the OEM extensions to the mps tables
> >> (if present). This gives me a mapping to tell which device
> >> lies in which NUMA node (the current code just guesses).
> >
> > So these extensions are OEM-specific, not part of the MP spec,
> > right?
> 
> As I understand this, the concept of OEM extensions is inside
> the MPS spec (there's a pointer for it inside the main table),
> though what's actually contained therein is OEM specific.

Right/agreed.  I didn't mean to imply that OEM extensions were
new, just that their contents are OEM-specific and not part
of the spec.

> > Also, could the array of structs <mp_irqs and mp_ioapics> (in
> > mpparse.c) be made __initdata, so that they could be discarded
> > after init?
> 
> Probably, but they're used in lots of places, so it would take some
> research to figure out all the possible combinations ;-) I'll leave that
> possiblity for a seperate patch (the structures were there already
> like this).

Right. :)

> A patch is attached below to fix these issues, plus one other bit
> of idiocy I found - I'd accidentally reduced the number of
> MAX_IRQ_SOURCES (I think I pulled the change forward from
> an older kernel, and dropped someone's fix. Oops).

Looks sane to me.

Thanks,
~Randy