Re: [PATCH 2/2] target/ppc/cpu_init: Synchronize HASHKEYR with KVM for migration

[Nicholas Piggin]

On Mon Jun 3, 2024 at 9:53 PM AEST, Shivaprasad G Bhat wrote:
> The patch enables HASHKEYR migration by hooking with the
> "KVM one reg" ID KVM_REG_PPC_HASHKEYR.
>
> Signed-off-by: Shivaprasad G Bhat <sbhat@linux.ibm.com>
> ---
>  linux-headers/asm-powerpc/kvm.h |    1 +
>  target/ppc/cpu_init.c           |    4 ++--
>  2 files changed, 3 insertions(+), 2 deletions(-)
>
> diff --git a/linux-headers/asm-powerpc/kvm.h b/linux-headers/asm-powerpc/kvm.h
> index fcb947f656..23a0af739c 100644
> --- a/linux-headers/asm-powerpc/kvm.h
> +++ b/linux-headers/asm-powerpc/kvm.h
> @@ -646,6 +646,7 @@ struct kvm_ppc_cpu_char {
>  #define KVM_REG_PPC_DAWR1    (KVM_REG_PPC | KVM_REG_SIZE_U64 | 0xc4)
>  #define KVM_REG_PPC_DAWRX1   (KVM_REG_PPC | KVM_REG_SIZE_U64 | 0xc5)
>  #define KVM_REG_PPC_DEXCR    (KVM_REG_PPC | KVM_REG_SIZE_U64 | 0xc6)
> +#define KVM_REG_PPC_HASHKEYR (KVM_REG_PPC | KVM_REG_SIZE_U64 | 0xc7)
>  
>  /* Transactional Memory checkpointed state:
>   * This is all GPRs, all VSX regs and a subset of SPRs
> diff --git a/target/ppc/cpu_init.c b/target/ppc/cpu_init.c
> index b1422c2eab..cee0a609eb 100644
> --- a/target/ppc/cpu_init.c
> +++ b/target/ppc/cpu_init.c
> @@ -5805,10 +5805,10 @@ static void register_power10_hash_sprs(CPUPPCState 
> *env)
>          ((uint64_t)g_rand_int(rand) << 32) | (uint64_t)g_rand_int(rand);
>      g_rand_free(rand);
>  #endif
> -    spr_register(env, SPR_HASHKEYR, "HASHKEYR",
> +    spr_register_kvm(env, SPR_HASHKEYR, "HASHKEYR",
>              SPR_NOACCESS, SPR_NOACCESS,
>              &spr_read_generic, &spr_write_generic,
> -            hashkeyr_initial_value);
> +            KVM_REG_PPC_HASHKEYR, hashkeyr_initial_value);
>      spr_register_hv(env, SPR_HASHPKEYR, "HASHPKEYR",
>              SPR_NOACCESS, SPR_NOACCESS,
>              SPR_NOACCESS, SPR_NOACCESS,

Hmm... now that I look at it, the hashpkey value also needs to be set
in the machine and migrated, right? That looks broken. I *think* if we
make this spr_register_kvm_hv, and you will also need to add a KVM
API for the register, that should get it working becuse SPRs will
be migrated for us.

Thanks,
Nick