On Wed, Oct 20, 2021, Paolo Bonzini wrote:

The kvm_x86_sync_pir_to_irr callback can sometimes set KVM_REQ_EVENT. If that happens exactly at the time that an exit is handled as EXIT_FASTPATH_REENTER_GUEST, vcpu_enter_guest will go incorrectly through the loop that calls kvm_x86_run, instead of processing the request promptly.

Fixes: 379a3c8ee444 ("KVM: VMX: Optimize posted-interrupt delivery for timer fastpath") Cc: stable@vger.kernel.org Signed-off-by: Paolo Bonzini pbonzini@redhat.com

---

Reviewed-by: Sean Christopherson seanjc@google.com

arch/x86/kvm/x86.c | 10 +++++----- 1 file changed, 5 insertions(+), 5 deletions(-)

diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c index fa48948b4934..b9b31e5f72b0 100644 --- a/arch/x86/kvm/x86.c +++ b/arch/x86/kvm/x86.c @@ -9781,14 +9781,14 @@ static int vcpu_enter_guest(struct kvm_vcpu *vcpu) if (likely(exit_fastpath != EXIT_FASTPATH_REENTER_GUEST)) break;

•            if (unlikely(kvm_vcpu_exit_request(vcpu))) {
  

• if (vcpu->arch.apicv_active)
  

• 	static_call(kvm_x86_sync_pir_to_irr)(vcpu);
  

• if (unlikely(kvm_vcpu_exit_request(vcpu))) {
  exit_fastpath = EXIT_FASTPATH_EXIT_HANDLED;
  break;
  

  }

• if (vcpu->arch.apicv_active)
  

• 	static_call(kvm_x86_sync_pir_to_irr)(vcpu);
  

•    }
  

• }

I think someone working on git has a meta-entry in the obfuscated C context. This is the most convoluted diff possible for a simple code move :-)

/* * Do this here before restoring debug registers on the host. And -- 2.27.0