Re: [PATCH 3/6] KVM: PPC: Book3E: Increase FPU laziness

[Scott Wood <scottwood@freescale.com>]

On 07/03/2013 12:23:16 PM, Alexander Graf wrote:
>=20
> On 03.07.2013, at 19:18, Scott Wood wrote:
>=20
> > On 07/03/2013 07:42:36 AM, Mihai Caraman wrote:
> >> Increase FPU laziness by calling kvmppc_load_guest_fp() just before
> >> returning to guest instead of each sched in. Without this =20
> improvement
> >> an interrupt may also claim floting point corrupting guest state.
> >> Signed-off-by: Mihai Caraman <mihai.caraman@freescale.com>
> >> ---
> >> arch/powerpc/kvm/booke.c  |    1 +
> >> arch/powerpc/kvm/e500mc.c |    2 --
> >> 2 files changed, 1 insertions(+), 2 deletions(-)
> >> diff --git a/arch/powerpc/kvm/booke.c b/arch/powerpc/kvm/booke.c
> >> index 113961f..3cae2e3 100644
> >> --- a/arch/powerpc/kvm/booke.c
> >> +++ b/arch/powerpc/kvm/booke.c
> >> @@ -1204,6 +1204,7 @@ int kvmppc_handle_exit(struct kvm_run *run, =20
> struct kvm_vcpu *vcpu,
> >> 			r =3D (s << 2) | RESUME_HOST | (r & =20
> RESUME_FLAG_NV);
> >> 		} else {
> >> 			kvmppc_lazy_ee_enable();
> >> +			kvmppc_load_guest_fp(vcpu);
> >> 		}
> >
> > This should go before the kvmppc_lazy_ee_enable().
>=20
> Why? What difference does that make? We're running with interrupts =20
> disabled here, right?

Yes, and we want to minimize the code we run where we have interrupts =20
disabled but the lazy ee state says they're enabled.  So =20
kvmppc_lazy_ee_enable() should be the last thing we do before entering =20
asm code.

See http://patchwork.ozlabs.org/patch/249565/

-Scott=