On 08/12/20 18:08, Maxim Levitsky wrote:
Even if you support TSCADJUST and let the guest write to it does not change the per guest offset at all. TSCADJUST is per [v]CPU and adds on top:
tscvcpu = tsc_host + guest_offset + TSC_ADJUSTScaling is just orthogonal and does not change any of this.
I agree with this, and I think that this is what we will end up doing. Paulo, what do you think about this?
Yes, you can have a VM ioctl that saves/restores cur_tsc_nsec and cur_tsc_write. The restore side would loop on all vcpus.
However, it is not so easy: 1) it would have to be usable only if KVM_X86_QUIRK_TSC_HOST_ACCESS is false, 2) it would fail if kvm->arch.nr_vcpus_matched_tsc == kvm->online_vcpus (which basically means that userspace didn't mess up the TSC configuration). If not, it would return -EINVAL.
Also, while at it let's burn and pour salt on the support for KVM_SET_TSC_KHZ unless TSC scaling is supported, together with vcpu->tsc_catchup and all the "tolerance" crap that is in kvm_set_tsc_khz. And initialize vcpu->arch.virtual_tsc_khz to kvm->arch.last_tsc_khz before calling kvm_synchronize_tsc.
Paolo