Re: [PATCH v4 05/15] KVM: s390: enable/disable AP interpretive execution

[Tony Krowiak <akrowiak@linux.vnet.ibm.com>]

On 04/16/2018 07:52 AM, Halil Pasic wrote:
>
> On 04/16/2018 01:13 PM, Pierre Morel wrote:
>> On 16/04/2018 12:51, Pierre Morel wrote:
>>> On 15/04/2018 23:22, Tony Krowiak wrote:
>>>> The VFIO AP device model exploits interpretive execution of AP
>>>> instructions (APIE) to provide guests passthrough access to AP
>>>> devices. This patch introduces a new interface to enable and
>>>> disable APIE.
>>>>
>>>> Signed-off-by: Tony Krowiak <akrowiak@linux.vnet.ibm.com>
>>>> ---
>>>>    arch/s390/include/asm/kvm-ap.h   |   16 ++++++++++++++++
>>>>    arch/s390/include/asm/kvm_host.h |    1 +
>>>>    arch/s390/kvm/kvm-ap.c           |   20 ++++++++++++++++++++
>>>>    arch/s390/kvm/kvm-s390.c         |    9 +++++++++
>>>>    4 files changed, 46 insertions(+), 0 deletions(-)
>>>>
>>>> diff --git a/arch/s390/include/asm/kvm-ap.h b/arch/s390/include/asm/kvm-ap.h
>>>> index 736e93e..a6c092e 100644
>>>> --- a/arch/s390/include/asm/kvm-ap.h
>>>> +++ b/arch/s390/include/asm/kvm-ap.h
>>>> @@ -35,4 +35,20 @@
>>>>     */
>>>>    void kvm_ap_build_crycbd(struct kvm *kvm);
>>>>
>>>> +/**
>>>> + * kvm_ap_interpret_instructions
>>>> + *
>>>> + * Indicate whether AP instructions shall be interpreted. If they are not
>>>> + * interpreted, all AP instructions will be intercepted and routed back to
>>>> + * userspace.
>>>> + *
>>>> + * @kvm: the virtual machine attributes
>>>> + * @enable: indicates whether AP instructions are to be interpreted (true) or
>>>> + *        or not (false).
>>>> + *
>>>> + * Returns 0 if completed successfully; otherwise, returns -EOPNOTSUPP
>>>> + * indicating that AP instructions are not installed on the guest.
>>>> + */
>>>> +int kvm_ap_interpret_instructions(struct kvm *kvm, bool enable);
>>>> +
>>>>    #endif /* _ASM_KVM_AP */
>>>> diff --git a/arch/s390/include/asm/kvm_host.h b/arch/s390/include/asm/kvm_host.h
>>>> index 3162783..5470685 100644
>>>> --- a/arch/s390/include/asm/kvm_host.h
>>>> +++ b/arch/s390/include/asm/kvm_host.h
>>>> @@ -715,6 +715,7 @@ struct kvm_s390_crypto {
>>>>        __u32 crycbd;
>>>>        __u8 aes_kw;
>>>>        __u8 dea_kw;
>>>> +    __u8 apie;
>>>>    };
>>>>
>>>>    #define APCB0_MASK_SIZE 1
>>>> diff --git a/arch/s390/kvm/kvm-ap.c b/arch/s390/kvm/kvm-ap.c
>>>> index 991bae4..55d11b5 100644
>>>> --- a/arch/s390/kvm/kvm-ap.c
>>>> +++ b/arch/s390/kvm/kvm-ap.c
>>>> @@ -58,3 +58,23 @@ void kvm_ap_build_crycbd(struct kvm *kvm)
>>>>        }
>>>>    }
>>>>    EXPORT_SYMBOL(kvm_ap_build_crycbd);
>>>> +
>>>> +int kvm_ap_interpret_instructions(struct kvm *kvm, bool enable)
>>>> +{
>>>> +    int ret = 0;
>>>> +
>>>> +    mutex_lock(&kvm->lock);
>>>> +
>>>> +    if (!test_kvm_cpu_feat(kvm, KVM_S390_VM_CPU_FEAT_AP)) {
>>> Do we really need to test CPU_FEAT_AP?
>>>
>>> I understand that KVM_S390_VM_CPU_FEAT_AP means AP instructions are interpreted.
>>> shouldn't we add this information in the name?
>>> like KVM_S390_VM_CPU_FEAT_APIE
>> If I misunderstood and FEAT_AP really mean AP instructions available in the guest,
>> same question:
>> is this function called if AP instructions are not available in the guest?
>>
> See patch #13. I guess the check above is anyway good as defensive
> programming. This implementation should be sane regardless of
> the answer to your question.

I agree.

>
>>>> +        ret = -EOPNOTSUPP;
>>>> +        goto done;
>>>> +    }
>>>> +
>>>> +    kvm->arch.crypto.apie = enable;
>>>> +    kvm_s390_vcpu_crypto_reset_all(kvm);
>>>> +
>>>> +done:
>>>> +    mutex_unlock(&kvm->lock);
>>>> +    return ret;
>>>> +}
>>>> +EXPORT_SYMBOL(kvm_ap_interpret_instructions);
>>>> diff --git a/arch/s390/kvm/kvm-s390.c b/arch/s390/kvm/kvm-s390.c
>>>> index 55cd897..1dc8566 100644
>>>> --- a/arch/s390/kvm/kvm-s390.c
>>>> +++ b/arch/s390/kvm/kvm-s390.c
>>>> @@ -1901,6 +1901,9 @@ static void kvm_s390_crypto_init(struct kvm *kvm)
>>>>        kvm->arch.crypto.crycb = &kvm->arch.sie_page2->crycb;
>>>>        kvm_ap_build_crycbd(kvm);
>>>>
>>>> +    /* Default setting indicating SIE shall interpret AP instructions */
>>>> +    kvm->arch.crypto.apie = 1;
>>>> +
>>>>        if (!test_kvm_facility(kvm, 76))
>>>>            return;
>>>>
>>>> @@ -2434,6 +2437,12 @@ static void kvm_s390_vcpu_crypto_setup(struct kvm_vcpu *vcpu)
>>>>    {
>>>>        vcpu->arch.sie_block->crycbd = vcpu->kvm->arch.crypto.crycbd;
>>>>
>>>> +    vcpu->arch.sie_block->eca &= ~ECA_APIE;
>>>> +    if (vcpu->kvm->arch.crypto.apie &&
>>>> +        test_kvm_cpu_feat(vcpu->kvm, KVM_S390_VM_CPU_FEAT_AP))
>>> Do we call xxx_crypto_setup() if KVM does not support AP interpretation?
>> sorry, I should have written AP instructions here:
>> is this function called if AP instructions are not available in the guest?
>>
> Yes, this function can be called with AP instructions available to the guest.
> Please have a look at patch 2 (kvm_s390_vm_set_crypto and the rest).
>
> Also this function is called on initialization regardless of AP instructions.
>
>>>> +        vcpu->arch.sie_block->eca |= ECA_APIE;
>>>> +
>>>> +
>>>>        if (!test_kvm_facility(vcpu->kvm, 76))
>>>>            return;
>>>>