Re: [RFC PATCH RESEND 3/4] target/arm/kvm: Handle KVM Target Imp CPU hypercalls

[Sebastian Ott <sebott@redhat.com>]

On Fri, 1 Aug 2025, Shameer Kolothum wrote:
> From: Shameer Kolothum <shameerali.kolothum.thodi@huawei.com>
>
> When target implementation CPUs are set, handle the related hyper calls
> correctly by returning the information requested.
>
> Signed-off-by: Shameer Kolothum <shameerali.kolothum.thodi@huawei.com>
> ---
> target/arm/kvm.c | 99 ++++++++++++++++++++++++++++++++++++++++++++++++
> 1 file changed, 99 insertions(+)
>
> diff --git a/target/arm/kvm.c b/target/arm/kvm.c
> index 8f325c4ca4..5adecc864e 100644
> --- a/target/arm/kvm.c
> +++ b/target/arm/kvm.c
> @@ -1697,6 +1697,71 @@ static bool kvm_arm_handle_debug(ARMCPU *cpu,
>     return false;
> }
>
> +/* Only supports HYP_KVM_DISCOVER_IMPL_XXX hypercalls */
> +static void arm_handle_smccc_kvm_vendor_hypercall(ARMCPU *cpu)
> +{
> +    CPUARMState *env = &cpu->env;
> +    uint64_t param[4] = { };
> +    uint64_t idx;
> +
> +    if (!is_a64(env)) {
> +        env->regs[0] = SMCCC_RET_NOT_SUPPORTED;
> +        return;
> +    }
> +
> +    memcpy(param, env->xregs, sizeof(param));
> +
> +    switch (param[0]) {
> +    case ARM_SMCCC_VENDOR_HYP_KVM_DISCOVER_IMPL_VER_FUNC_ID:
> +        if (!target_impl_cpus_num) {
> +            env->xregs[0] = SMCCC_RET_NOT_SUPPORTED;
> +            return;
> +        }
> +        env->xregs[0] = SMCCC_RET_SUCCESS;
> +        env->xregs[1] = PSCI_VERSION(1, 0);
> +        env->xregs[2] = target_impl_cpus_num;
> +        break;
> +    case ARM_SMCCC_VENDOR_HYP_KVM_DISCOVER_IMPL_CPUS_FUNC_ID:
> +        idx = param[1];
> +
> +        if (!target_impl_cpus_num || idx >= target_impl_cpus_num) {
> +            env->xregs[0] = SMCCC_RET_INVALID_PARAMETER;
> +            return;
> +        }
> +
> +        env->xregs[0] = SMCCC_RET_SUCCESS;
> +        env->xregs[1] = target_impl_cpus[idx].midr;
> +        env->xregs[2] = target_impl_cpus[idx].revidr;
> +        env->xregs[3] = target_impl_cpus[idx].aidr;
> +        break;
> +    default:
> +        env->xregs[0] = SMCCC_RET_NOT_SUPPORTED;
> +    }
> +}
> +
> +static int kvm_arm_handle_hypercall(CPUState *cs, struct kvm_run *run)
> +{
> +    ARMCPU *cpu = ARM_CPU(cs);
> +    CPUARMState *env = &cpu->env;
> +
> +    kvm_cpu_synchronize_state(cs);
> +
> +    if (run->hypercall.flags == KVM_HYPERCALL_EXIT_SMC) {
> +        cs->exception_index = EXCP_SMC;
> +        env->exception.syndrome = syn_aa64_smc(0);
> +    } else {
> +        cs->exception_index = EXCP_HVC;
> +        env->exception.syndrome = syn_aa64_hvc(0);
> +    }
> +    env->exception.target_el = 1;
> +
> +    bql_lock();
> +    arm_handle_smccc_kvm_vendor_hypercall(cpu);
> +    bql_unlock();
> +
> +    return EXCP_INTERRUPT;
> +}
> +
> int kvm_arch_handle_exit(CPUState *cs, struct kvm_run *run)
> {
>     ARMCPU *cpu = ARM_CPU(cs);
> @@ -1713,6 +1778,9 @@ int kvm_arch_handle_exit(CPUState *cs, struct kvm_run *run)
>         ret = kvm_arm_handle_dabt_nisv(cpu, run->arm_nisv.esr_iss,
>                                        run->arm_nisv.fault_ipa);
>         break;
> +    case KVM_EXIT_HYPERCALL:
> +        ret = kvm_arm_handle_hypercall(cs, run);
> +        break;
>     default:
>         qemu_log_mask(LOG_UNIMP, "%s: un-handled exit reason %d\n",
>                       __func__, run->exit_reason);
> @@ -2212,6 +2280,29 @@ static int kvm_arm_sve_set_vls(ARMCPU *cpu)
>     return kvm_set_one_reg(CPU(cpu), KVM_REG_ARM64_SVE_VLS, &vls[0]);
> }
>
> +/*
> + * Supported Target Implementation CPU hypercalls:
> + *   KVM_REG_ARM_VENDOR_HYP_BIT_DISCOVER_IMPL_VER   = 0,
> + *   KVM_REG_ARM_VENDOR_HYP_BIT_DISCOVER_IMPL_CPUS  = 1
> + *
> + * Setting these bits advertises the availability of the corresponding
> + * Target Implementation CPU hypercalls to the guest.
> + */
> +#define BMAP_2_DISCOVER_IMPL_BITS 0x3ULL
> +static int kvm_arm_target_impl_cpus_set_hyp_bmap2(ARMCPU *cpu)
> +{
> +    uint64_t bmap2;
> +    int ret;
> +
> +    ret = kvm_get_one_reg(CPU(cpu), KVM_REG_ARM_VENDOR_HYP_BMAP_2, &bmap2);
> +    if (ret) {
> +        return ret;
> +    }
> +
> +    bmap2 |= BMAP_2_DISCOVER_IMPL_BITS;
> +    return kvm_set_one_reg(CPU(cpu), KVM_REG_ARM_VENDOR_HYP_BMAP_2, &bmap2);
> +}
> +
> #define ARM_CPU_ID_MPIDR       3, 0, 0, 0, 5
>
> int kvm_arch_init_vcpu(CPUState *cs)
> @@ -2293,6 +2384,14 @@ int kvm_arch_init_vcpu(CPUState *cs)
>     }
>     cpu->mp_affinity = mpidr & ARM64_AFFINITY_MASK;
>
> +    /* Set KVM_REG_ARM_VENDOR_HYP_BMAP_2 if target impl CPUs are required */
> +    if (target_impl_cpus_num) {
> +        ret = kvm_arm_target_impl_cpus_set_hyp_bmap2(cpu);

There's still the question of what we use to actually set the
imp id regs for the vm. I'm in favour of using target_impl_cpus[0].*
but if these can also be specified via the writable id regs interface
(https://lore.kernel.org/qemu-devel/20250414163849.321857-1-cohuck@redhat.com/)
we need to make sure that they match.

Otherwise, looks good to me!

Sebastian