Re: [Qemu-ppc] [PATCH v2] spapr: Don't set the TM ibm, pa-features bit in PR KVM mode

[David Gibson]

On Sat, Apr 30, 2016 at 10:48:00AM +1000, Anton Blanchard wrote:
> We don't support transactional memory in PR KVM, so don't tell
> the OS that we do.
> 
> Signed-off-by: Anton Blanchard <address@hidden>

Relying on CAP_PPC_GET_PVINFO is a hack we need in some cases, but
it's not something to be encouraged.  I'd prefer to see this examining
a new capability specifically advertising TM support in KVM, with a
fallback to PVINFO if necessary.

Come to that, we probably shouldn't be advertising TM support in TCG,
either, since IIRC our "support" for the TM instructions there is
basically worse than useless.

> ---
> 
> v2: Fix build with CONFIG_KVM disabled, noticed by Alex.
> 
> diff --git a/hw/ppc/spapr.c b/hw/ppc/spapr.c
> index b69995e..dc3e3c9 100644
> --- a/hw/ppc/spapr.c
> +++ b/hw/ppc/spapr.c
> @@ -696,6 +696,14 @@ static void spapr_populate_cpu_dt(CPUState *cs, void 
> *fdt, int offset,
>      } else /* env->mmu_model == POWERPC_MMU_2_07 */ {
>          pa_features = pa_features_207;
>          pa_size = sizeof(pa_features_207);
> +
> +#ifdef CONFIG_KVM
> +        /* Don't enable TM in PR KVM mode */
> +        if (kvm_enabled() &&
> +            kvm_vm_check_extension(cs->kvm_state, KVM_CAP_PPC_GET_PVINFO)) {
> +            pa_features[24] &= ~0x80;
> +        }
> +#endif
>      }
>      if (env->ci_large_pages) {
>          pa_features[3] |= 0x20;
> 

-- 
David Gibson                    | I'll have my music baroque, and my code
david AT gibson.dropbear.id.au  | minimalist, thank you.  NOT _the_ _other_
                                | _way_ _around_!
http://www.ozlabs.org/~dgibson

signature.asc
Description: PGP signature