Re: runtime regression with "x86/mm/pat: Emulate PAT when it is disabled"

From: Paul Gortmaker
Date: Thu Mar 10 2016 - 14:04:56 EST

[Re: runtime regression with "x86/mm/pat: Emulate PAT when it is disabled"] On 10/03/2016 (Thu 18:20) Borislav Petkov wrote:

> On Thu, Mar 10, 2016 at 09:49:51AM -0700, Toshi Kani wrote:
> > This confirms the issue - QEMU's virtual Intel CPU does not support MTRR. 
> >
> > When MTRR is disabled, the kernel does not call pat_init().  pat_enabled()
> > is still set to true when CONFIG_X86_PAT is set.  CONFIG_X86_PAT depends on
> > CONFIG_MTRR, and assumes that MTRR is enabled. 
> Aha, so "qemu32" model doesn't support MTRRs but "kvm32" does, for
> example. And so do the majority of the other CPU types.

So, I guess that is a qemu bug? If there is no real silicon out there
that has no MTRR but does claim PAT, then qemu32 is a flawed CPU type?

> Paul, can you guys run with something else besides "qemu32"? You can
> even take a 64-bit one and run a 32-bit guest on it.

That is probably more of an RP question. In principle I guess other CPU
types are on the table, and most likely qemu32 is just there from
historical reasons. We do know that we don't want "-cpu host" though,
since that will introduce variability into the automated testing.


> :-)
> --
> Regards/Gruss,
> Boris.
> SUSE Linux GmbH, GF: Felix Imendörffer, Jane Smithard, Graham Norton, HRB 21284 (AG Nürnberg)
> --