releases/4.19.35/kvm-nvmx-nmi-window-and-interrupt-window-exiting-sho.patch

   1 From 0b565f2d6cd76df33c6658b01a45ae36a4404826 Mon Sep 17 00:00:00 2001
   2 From: Jim Mattson <jmattson@google.com>
   3 Date: Mon, 26 Nov 2018 11:22:32 -0800
   4 Subject: kvm: nVMX: NMI-window and interrupt-window exiting should wake L2
   5  from HLT
   6 MIME-Version: 1.0
   7 Content-Type: text/plain; charset=UTF-8
   8 Content-Transfer-Encoding: 8bit
   9
  10 [ Upstream commit 9ebdfe5230f2e50e3ba05c57723a06e90946815a ]
  11
  12 According to the SDM, "NMI-window exiting" VM-exits wake a logical
  13 processor from the same inactive states as would an NMI and
  14 "interrupt-window exiting" VM-exits wake a logical processor from the
  15 same inactive states as would an external interrupt. Specifically, they
  16 wake a logical processor from the shutdown state and from the states
  17 entered using the HLT and MWAIT instructions.
  18
  19 Fixes: 6dfacadd5858 ("KVM: nVMX: Add support for activity state HLT")
  20 Signed-off-by: Jim Mattson <jmattson@google.com>
  21 Reviewed-by: Peter Shier <pshier@google.com>
  22 Suggested-by: Sean Christopherson <sean.j.christopherson@intel.com>
  23 [Squashed comments of two Jim's patches and used the simplified code
  24  hunk provided by Sean. - Radim]
  25 Signed-off-by: Radim Krčmář <rkrcmar@redhat.com>
  26 Signed-off-by: Sasha Levin <sashal@kernel.org>
  27 ---
  28  arch/x86/kvm/vmx.c | 10 +++++++---
  29  1 file changed, 7 insertions(+), 3 deletions(-)
  30
  31 diff --git a/arch/x86/kvm/vmx.c b/arch/x86/kvm/vmx.c
  32 index f99f59625da5..e61ac229a6c1 100644
  33 --- a/arch/x86/kvm/vmx.c
  34 +++ b/arch/x86/kvm/vmx.c
  35 @@ -12836,11 +12836,15 @@ static int nested_vmx_run(struct kvm_vcpu *vcpu, bool launch)
  36         nested_cache_shadow_vmcs12(vcpu, vmcs12);
  37
  38         /*
  39 -        * If we're entering a halted L2 vcpu and the L2 vcpu won't be woken
  40 -        * by event injection, halt vcpu.
  41 +        * If we're entering a halted L2 vcpu and the L2 vcpu won't be
  42 +        * awakened by event injection or by an NMI-window VM-exit or
  43 +        * by an interrupt-window VM-exit, halt the vcpu.
  44          */
  45         if ((vmcs12->guest_activity_state == GUEST_ACTIVITY_HLT) &&
  46 -           !(vmcs12->vm_entry_intr_info_field & INTR_INFO_VALID_MASK)) {
  47 +           !(vmcs12->vm_entry_intr_info_field & INTR_INFO_VALID_MASK) &&
  48 +           !(vmcs12->cpu_based_vm_exec_control & CPU_BASED_VIRTUAL_NMI_PENDING) &&
  49 +           !((vmcs12->cpu_based_vm_exec_control & CPU_BASED_VIRTUAL_INTR_PENDING) &&
  50 +             (vmcs12->guest_rflags & X86_EFLAGS_IF))) {
  51                 vmx->nested.nested_run_pending = 0;
  52                 return kvm_vcpu_halt(vcpu);
  53         }
  54 --
  55 2.19.1
  56