Re: [PATCH] Raise maximum number of memory controllers

[Luck, Tony]

On Tue, Sep 25, 2018 at 05:26:59PM +0200, Borislav Petkov wrote:
> On Tue, Sep 25, 2018 at 09:34:49AM -0500, Justin Ernst wrote:
> > We observe an oops in the skx_edac module during boot.
> > Examining /var/log/messages:
> > [ 3401.985757] EDAC MC0: Giving out device to module skx_edac controller Skylake Socket#0 IMC#0
> > [ 3401.985887] EDAC MC1: Giving out device to module skx_edac controller Skylake Socket#0 IMC#1
> > [ 3401.986014] EDAC MC2: Giving out device to module skx_edac controller Skylake Socket#1 IMC#0
> > ...
> > [ 3401.987318] EDAC MC13: Giving out device to module skx_edac controller Skylake Socket#0 IMC#1
> > [ 3401.987435] EDAC MC14: Giving out device to module skx_edac controller Skylake Socket#1 IMC#0
> > [ 3401.987556] EDAC MC15: Giving out device to module skx_edac controller Skylake Socket#1 IMC#1
> > [ 3401.987579] Too many memory controllers: 16
> > [ 3402.042614] EDAC MC: Removed device 0 for skx_edac Skylake Socket#0 IMC#0
> > 
> > We observe there are two memory controllers per socket, with a limit of 16.
> > Raise the maximum number of memory controllers from 16 to 2 * MAX_NUMNODES (1024).
> 
> Tony,
> 
> can we read that out from the hardware instead of having this silly
> static number?
> 
> Leaving in the rest.

There are way too many places where we use the identifier "bus"
in the edac core and drivers. But I'm not sure that we need a
static array mc_bus[EDAC_MAX_MCS].

Why can't we:


-	mci->bus = &mc_bus[mci->mc_idx];
+	mci->bus = kmalloc(sizeof *(mci->bus), GFP_KERNEL);

and then figure out where to kfree(mci->bus) on driver removal?

Do we every do arithmetic on different mci->bus pointers that
assume they are all part of a single array?

-Tony