Hi Jörg, On Wed, Sep 16, 2026 at 06:31:04PM +0200, Jörg Rödel wrote: > On Tue, Sep 15, 2026 at 02:30:36PM +0300, Tony Lindgren wrote: > > Nice, looks like snp_get/set_vcpu_state() could quite easily make use > > of KVM_EXPORT/IMPORT_VCPU? > > That depends on whether KVM_EXPORT/IMPORT_VCPU will migrate the confidential > VCPU state (registers) only or hypervisor state as well. The get/set_state > helpers here only export/import the hypervisor tracking state for the guest > VCPUs.
Oh OK, so calling for both hypervisor and confidential VCPU state would be needed. The difference could be specified in the struct kvm_vcpu_transfer flags? > On AMD the confidential VCPU state is migrated as part of the memory image > when > guest execution has stopped on the source side. This part for the confidential VCPU transfer matches what we are doing for TDX. Regards, Tony
