On 2/24/25 18:54, Sean Christopherson wrote:
> On Mon, Feb 24, 2025, Tom Lendacky wrote:
>> On 2/24/25 16:55, Sean Christopherson wrote:
>>> On Mon, Feb 24, 2025, Tom Lendacky wrote:
>>>> On 2/18/25 19:26, Sean Christopherson wrote:
>
> Given that, IIUC, KVM would eventually return KVM_EXIT_FAIL_ENTRY, I like your
> idea of returning meaningful information. And unless I'm missing something, that
> would obviate any need to terminate the VM, which would address your earlier point
> of whether terminating the VM is truly better than returning than returning a
> familiar error code.
>
> So this? (completely untested)
This works nicely and qemu terminates quickly with:
KVM: entry failed, hardware error 0xffffffffffffffff
Thanks,
Tom
>
> diff --git a/arch/x86/kvm/svm/sev.c b/arch/x86/kvm/svm/sev.c
> index 7345cac6f93a..71b340cbe561 100644
> --- a/arch/x86/kvm/svm/sev.c
> +++ b/arch/x86/kvm/svm/sev.c
> @@ -3463,10 +3463,8 @@ int pre_sev_run(struct vcpu_svm *svm, int cpu)
> * invalid VMSA, e.g. if userspace forces the vCPU to be RUNNABLE after
> * an SNP AP Destroy event.
> */
> - if (sev_es_guest(kvm) && !VALID_PAGE(svm->vmcb->control.vmsa_pa)) {
> - kvm_vm_dead(kvm);
> - return -EIO;
> - }
> + if (sev_es_guest(kvm) && !VALID_PAGE(svm->vmcb->control.vmsa_pa))
> + return -EINVAL;
>
> /* Assign the asid allocated with this SEV guest */
> svm->asid = asid;
> diff --git a/arch/x86/kvm/svm/svm.c b/arch/x86/kvm/svm/svm.c
> index 46e0b65a9fec..f72bcf2e590e 100644
> --- a/arch/x86/kvm/svm/svm.c
> +++ b/arch/x86/kvm/svm/svm.c
> @@ -4233,8 +4233,12 @@ static __no_kcsan fastpath_t svm_vcpu_run(struct kvm_vcpu *vcpu,
> if (force_immediate_exit)
> smp_send_reschedule(vcpu->cpu);
>
> - if (pre_svm_run(vcpu))
> + if (pre_svm_run(vcpu)) {
> + vcpu->run->exit_reason = KVM_EXIT_FAIL_ENTRY;
> + vcpu->run->fail_entry.hardware_entry_failure_reason = SVM_EXIT_ERR;
> + vcpu->run->fail_entry.cpu = vcpu->cpu;
> return EXIT_FASTPATH_EXIT_USERSPACE;
> + }
>
> sync_lapic_to_cr8(vcpu);
