Re: [PATCH v13 13/48] kvm: arm64: Don't expose unsupported capabilities for realm guests

[Steven Price <steven.price@arm.com>]

On 19/03/2026 14:09, Suzuki K Poulose wrote:
> On 18/03/2026 15:53, Steven Price wrote:
>> From: Suzuki K Poulose <suzuki.poulose@arm.com>
>>
>> RMM v1.0 provides no mechanism for the host to perform debug operations
>> on the guest. So limit the extensions that are visible to an allowlist
>> so that only those capabilities we can support are advertised.
>>
>> Signed-off-by: Suzuki K Poulose <suzuki.poulose@arm.com>
>> Signed-off-by: Steven Price <steven.price@arm.com>
>> ---
>> Changes since v10:
>>   * Add a kvm_realm_ext_allowed() function which limits which extensions
>>     are exposed to an allowlist. This removes the need for special casing
>>     various extensions.
>> Changes since v7:
>>   * Remove the helper functions and inline the kvm_is_realm() check with
>>     a ternary operator.
>>   * Rewrite the commit message to explain this patch.
>> ---
>>   arch/arm64/kvm/arm.c | 22 ++++++++++++++++++++++
>>   1 file changed, 22 insertions(+)
>>
>> diff --git a/arch/arm64/kvm/arm.c b/arch/arm64/kvm/arm.c
>> index 9b17bdfaf0c2..ddbf080e4f55 100644
>> --- a/arch/arm64/kvm/arm.c
>> +++ b/arch/arm64/kvm/arm.c
>> @@ -357,6 +357,25 @@ static bool kvm_has_full_ptr_auth(void)
>>           (apa + api + apa3) == 1);
>>   }
>>   +static bool kvm_realm_ext_allowed(long ext)
>> +{
>> +    switch (ext) {
>> +    case KVM_CAP_IRQCHIP:
>> +    case KVM_CAP_ARM_PSCI:
>> +    case KVM_CAP_ARM_PSCI_0_2:
>> +    case KVM_CAP_NR_VCPUS:
>> +    case KVM_CAP_MAX_VCPUS:
>> +    case KVM_CAP_MAX_VCPU_ID:
>> +    case KVM_CAP_MSI_DEVID:
>> +    case KVM_CAP_ARM_VM_IPA_SIZE:
>> +    case KVM_CAP_ARM_PTRAUTH_ADDRESS:
>> +    case KVM_CAP_ARM_PTRAUTH_GENERIC:
>> +    case KVM_CAP_ARM_RMI:
>> +        return true;
>> +    }
>> +    return false;
>> +}
>> +
>>   int kvm_vm_ioctl_check_extension(struct kvm *kvm, long ext)
>>   {
>>       int r;
>> @@ -364,6 +383,9 @@ int kvm_vm_ioctl_check_extension(struct kvm *kvm,
>> long ext)
>>       if (is_protected_kvm_enabled() && !kvm_pkvm_ext_allowed(kvm, ext))
>>           return 0;
>>   +    if (kvm && kvm_is_realm(kvm) && !kvm_realm_ext_allowed(ext))
>> +        return 0;
>> +
> 
> We need a similar check in in kvm_vm_ioctl_enable_cap() to prevent
> enabling the filtered caps ? Otherwise looks good to me.

Indeed - thanks for spotting.

Thanks,
Steve

> Suzuki
> 
>>       switch (ext) {
>>       case KVM_CAP_IRQCHIP:
>>           r = vgic_present;
>