v1 -> v2:
1. In patch# 1, the invalid VM-entry MSR-load area for vmcs02 is
now a system-wide entity. It is allocated and initialized
during VMX initialization. The exit qualification is now
contained in a 32-bit variable in 'struct nested_vmx'.
2. Patch# 2 is new. It rolls back MSR updates when VM-entry
fails due to invalid VM-entry MSR-load area.
Some VM-entry checks can be offloaded from KVM to hardware. But if we want to
do that, the current implementation of KVM creates a priority issue where the
order in which VM-entry checks need to be performed according to the SDM, is
not maintained. VM-entry fails in nested_vmx_enter_non_root_mode() if an error
is encountered while processing the entries in VM-entry MSR-load area. This
leads to VM-exit due to a VM-entry check that is supposed to be done after
any guest-state checks done in hardware. This patch fixes this priority issue
so that checks that can be offloaded to hardware can now be offloaded.
[PATCH 1/2 v2] nVMX: Defer error from VM-entry MSR-load area to until
[PATCH 2/2 v2] nVMX: Rollback MSR-load if VM-entry fails due to VM-entry
arch/x86/kvm/vmx/nested.c | 84 +++++++++++++++++++++++++++++++++++++++++++----
arch/x86/kvm/vmx/nested.h | 29 ++++++++++++++--
arch/x86/kvm/vmx/vmx.c | 18 ++++++++++
arch/x86/kvm/vmx/vmx.h | 14 ++++++++
4 files changed, 136 insertions(+), 9 deletions(-)
Krish Sadhukhan (2):
nVMX: Defer error from VM-entry MSR-load area to until after hardware verifies VMCS guest state-area
nVMX: Rollback MSR-load if VM-entry fails due to VM-entry MSR-loading
