Fix two edge cases in nested VM-Enter where KVM loads garbage into
BNDCFGS and/or DEBUGCTL when migrating a vCPU with L2 active. If userspace
restores MSRs before nested state, userspace expects KVM to propagate L2's
values to vmcs02. KVM already mostly handles this scenario, so even though
I personally think it's rather ridiculous, it's easy for KVM to support,
and given that our VMM does KVM_SET_MSRS before KVM_SET_NESTED_STATE...
Patches 4 and 5 are minor optimizations to handle BNDCFGS more like
DEBUGCTL and/or the PKS MSR.
Sean Christopherson (5):
KVM: nVMX: Snapshot pre-VM-Enter BNDCFGS for !nested_run_pending case
KVM: nVMX: Snapshot pre-VM-Enter DEBUGCTL for !nested_run_pending case
KVM: nVMX: Rename nested.vmcs01_* fields to nested.pre_vmenter_*
KVM: nVMX: Save BNDCFGS to vmcs12 iff relevant controls are exposed to
L1
KVM: nVMX: Update vmcs12 on BNDCFGS write, not at vmcs02=>vmcs12 sync
arch/x86/kvm/vmx/nested.c | 16 ++++++++--------
arch/x86/kvm/vmx/vmx.c | 13 +++++++++++++
arch/x86/kvm/vmx/vmx.h | 15 ++++++++++++---
3 files changed, 33 insertions(+), 11 deletions(-)
base-commit: 8baacf67c76c560fed954ac972b63e6e59a6fba0
--
2.36.1.476.g0c4daa206d-goog
