Re: [RFC PATCH 4/4] spapr: Add KVM-on-TCG migration support

[Nicholas Piggin]

Excerpts from Fabiano Rosas's message of February 25, 2022 4:58 am:
> This adds migration support for TCG pseries machines running a KVM-HV
> guest.
> 
> The state that needs to be migrated is:
> 
> - the nested PTCR value;
> - the in_nested flag;
> - the nested_tb_offset.
> - the saved host CPUPPCState structure;
> 
> Signed-off-by: Fabiano Rosas <farosas@linux.ibm.com>

The series generally looks good to me, I guess patches 1 and 2 are
fixes that could go ahead.

Main thing about this is I was thinking of cutting down the CPUPPCState
structure for saving the host state when in the L2, and making a
specific structure for that that only contains what is required.

This patch could easily switch to that so it's no big deal AFAIKS.

> diff --git a/target/ppc/machine.c b/target/ppc/machine.c
> index 7ee1984500..ae09b1bcfe 100644
> --- a/target/ppc/machine.c
> +++ b/target/ppc/machine.c
> @@ -10,6 +10,7 @@
>  #include "kvm_ppc.h"
>  #include "power8-pmu.h"
>  #include "hw/ppc/ppc.h"
> +#include "hw/ppc/spapr_cpu_core.h"
>  
>  static void post_load_update_msr(CPUPPCState *env)
>  {
> @@ -679,6 +680,48 @@ static const VMStateDescription vmstate_tb_env = {
>      }
>  };
>  
> +static const VMStateDescription vmstate_hdecr = {
> +    .name = "cpu/hdecr",
> +    .version_id = 1,
> +    .minimum_version_id = 1,
> +    .fields = (VMStateField[]) {
> +        VMSTATE_UINT64(hdecr_next, ppc_tb_t),
> +        VMSTATE_TIMER_PTR(hdecr_timer, ppc_tb_t),
> +        VMSTATE_END_OF_LIST()
> +    }
> +};
> +
> +static bool nested_needed(void *opaque)
> +{
> +    PowerPCCPU *cpu = opaque;
> +    SpaprCpuState *spapr_cpu = spapr_cpu_state(cpu);
> +
> +    return spapr_cpu->in_nested;
> +}

I don't know the migration code -- are you assured of having a
spapr CPU here?

Maybe this could call a helper function located near the spapr/nested
code like 'return ppc_cpu_need_hdec_migrate(cpu)' ?

Thanks,
Nick