Hi, This series, of a single patch, aim to fix a bug of wrongly overwriting vmcs02->eoi_exit_bitmap when one L1 vCPU triggers an IOAPIC scan request while another L1 vCPU is running in guest-mode (L2). The patch fixes the issue by delaying the load of EOI-exitmap to happen only when vCPU is running L1. We have also written kvm-unit-tests that serves as a regression test to verify that this bug is indeed fixed. We will commit that kvm-unit-test in a seperate series of patches soon. In addition, we have written 2 kvm-unit-tests for the use-case of L1 pass-through it's LAPIC & IOAPIC to L2. As this patch is slightly related to that use-case as well. Changes from v1 to v2: * Delay only the load of EOI-exitmap instead of entire KVM_REQ_SCAN_IOAPIC which also recalc IOAPIC handled vectors. This was needed to correctly support use-case of L1 pass-through it's LAPIC & IOAPIC to L2. Regards, -Liran Alon (Previous version of patch can be found here: https://www.spinics.net/lists/kvm/msg158546.html)
