Re: [Qemu-ppc] [PATCH v7 5/6] ppc: spapr: Enable FWNMI capability

[David Gibson]

On Mon, Apr 15, 2019 at 03:42:46PM +0530, Aravinda Prasad wrote:
> 
> 
> On Tuesday 26 March 2019 05:03 AM, David Gibson wrote:
> > On Mon, Mar 25, 2019 at 02:27:45PM +0530, Aravinda Prasad wrote:
> >>
> >>
> >> On Monday 25 March 2019 12:02 PM, David Gibson wrote:
> >>> On Fri, Mar 22, 2019 at 12:04:16PM +0530, Aravinda Prasad wrote:
> >>>> Enable the KVM capability KVM_CAP_PPC_FWNMI so that
> >>>> the KVM causes guest exit with NMI as exit reason
> >>>> when it encounters a machine check exception on the
> >>>> address belonging to a guest. Without this capability
> >>>> enabled, KVM redirects machine check exceptions to
> >>>> guest's 0x200 vector.
> >>>>
> >>>> Signed-off-by: Aravinda Prasad <address@hidden>
> >>>> ---
> >>>>  hw/ppc/spapr_rtas.c  |   15 +++++++++++++++
> >>>>  target/ppc/kvm.c     |   14 ++++++++++++++
> >>>>  target/ppc/kvm_ppc.h |    6 ++++++
> >>>>  3 files changed, 35 insertions(+)
> >>>>
> >>>> diff --git a/hw/ppc/spapr_rtas.c b/hw/ppc/spapr_rtas.c
> >>>> index fb594a4..939f428 100644
> >>>> --- a/hw/ppc/spapr_rtas.c
> >>>> +++ b/hw/ppc/spapr_rtas.c
> >>>> @@ -49,6 +49,7 @@
> >>>>  #include "hw/ppc/fdt.h"
> >>>>  #include "target/ppc/mmu-hash64.h"
> >>>>  #include "target/ppc/mmu-book3s-v3.h"
> >>>> +#include "kvm_ppc.h"
> >>>>  
> >>>>  static void rtas_display_character(PowerPCCPU *cpu, SpaprMachineState 
> >>>> *spapr,
> >>>>                                     uint32_t token, uint32_t nargs,
> >>>> @@ -354,6 +355,20 @@ static void rtas_ibm_nmi_register(PowerPCCPU *cpu,
> >>>>                                    target_ulong args,
> >>>>                                    uint32_t nret, target_ulong rets)
> >>>>  {
> >>>> +    int ret;
> >>>> +
> >>>> +    ret = kvmppc_fwnmi_enable(cpu);
> >>>> +
> >>>> +    if (ret == 1) {
> >>>> +        rtas_st(rets, 0, RTAS_OUT_NOT_SUPPORTED);
> >>>
> >>> Urgh, we're here making a guest visible different to the environment
> >>> depending on a host (KVM) capability.  What happens if you start a
> >>> guest and it registers fwnmi support, then migrate it to a host that
> >>> lacks the necessary KVM support?
> >>
> >> I just checked how such scenarios are handled for other KVM
> >> capabilities. Should I need to add an Spapr cap for this?
> > 
> > Yes, I think that's what we'll need to do.
> 
> I was looking into SPAPR Cap, and I am not sure that the following code
> will help in handling the migration. I need some help here.
> 
> diff --git a/hw/ppc/spapr_caps.c b/hw/ppc/spapr_caps.c
> index edc5ed0..ef96192 100644
> --- a/hw/ppc/spapr_caps.c
> +++ b/hw/ppc/spapr_caps.c
> @@ -473,6 +473,15 @@ static void cap_ccf_assist_apply(SpaprMachineState
> *spapr, uint8_t val,
>      }
>  }
> 
> +static void cap_machine_check_apply(SpaprMachineState *spapr, uint8_t val,
> +                                    Error **errp)
> +{
> +    if (kvm_enabled()) {
> +        if (kvmppc_fwnmi_enable(cpu)) {
> +            error_setg(errp, "Unable to enable fwnmi capability");
> +    }
> +}
> +
>  SpaprCapabilityInfo capability_table[SPAPR_CAP_NUM] = {
>      [SPAPR_CAP_HTM] = {
>          .name = "htm",
> @@ -571,6 +580,15 @@ SpaprCapabilityInfo capability_table[SPAPR_CAP_NUM] = {
>          .type = "bool",
>          .apply = cap_ccf_assist_apply,
>      },
> +    [SPAPR_CAP_MACHINE_CHECK] = {
> +        .name = "machine-check",
> +        .description = "Handle machine check exceptions",
> +        .index = SPAPR_CAP_MACHINE_CHECK,
> +        .get = spapr_cap_get_bool,
> +        .set = spapr_cap_set_bool,
> +        .type = "bool",
> +        .apply = cap_machine_check_apply,
> +    },
>  };
> 
>  static SpaprCapabilities default_caps_with_cpu(SpaprMachineState *spapr,
> @@ -706,6 +724,7 @@ SPAPR_CAP_MIG_STATE(ibs, SPAPR_CAP_IBS);
>  SPAPR_CAP_MIG_STATE(nested_kvm_hv, SPAPR_CAP_NESTED_KVM_HV);
>  SPAPR_CAP_MIG_STATE(large_decr, SPAPR_CAP_LARGE_DECREMENTER);
>  SPAPR_CAP_MIG_STATE(ccf_assist, SPAPR_CAP_CCF_ASSIST);
> +SPAPR_CAP_MIG_STATE(mce, SPAPR_CAP_MACHINE_CHECK);
> 
>  void spapr_caps_init(SpaprMachineState *spapr)
>  {
> 
> 
> Or is it that just adding SPAPR_CAP_MIG_STATE(mce,
> SPAPR_CAP_MACHINE_CHECK); is enough as it is checking or cap values?

No, the change you have above looks mostly good.  You'll also need to
set default values for the cap in spapr_machine_class_init(), and
probably compat values in some of the earlier machine type variants.

I also don't like the cap name: there's at least some kind of handling
of machine checks before this, this is more specifically about the
fwnmi scheme, AIUI.

-- 
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