From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1757603AbcFAMfI (ORCPT ); Wed, 1 Jun 2016 08:35:08 -0400 Received: from mail-wm0-f68.google.com ([74.125.82.68]:35377 "EHLO mail-wm0-f68.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751969AbcFAMfF (ORCPT ); Wed, 1 Jun 2016 08:35:05 -0400 From: Paolo Bonzini To: linux-kernel@vger.kernel.org, kvm@vger.kernel.org Subject: [PATCH 0/2] KVM: x86: avoid simultaneous queueing of both IRQ and SMI Date: Wed, 1 Jun 2016 14:34:59 +0200 Message-Id: <1464784501-13710-1-git-send-email-pbonzini@redhat.com> X-Mailer: git-send-email 2.5.5 Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org This was reported as a vmentry failure while running Windows with SMM enabled. It's not that rare if your processor lacks APICv---it happens about 20-30% of the time while installing Windows 10. I now understand the interrupt injection code (especially complete_interrupts) better, and I also understand why the shortcut I took in SMI handling was a bad idea. In the end the code is somewhat simpler with the patch applied than before. The bug report is at https://github.com/tianocore/edk2/issues/91, but it also fixes other Windows failures that Laszlo had reported to me privately. Paolo Paolo Bonzini (2): KVM: x86: avoid simultaneous queueing of both IRQ and SMI KVM: x86: rename process_smi to enter_smm, process_smi_request to process_smi arch/x86/kvm/x86.c | 75 ++++++++++++++++++++++++++++++++---------------------- 1 file changed, 44 insertions(+), 31 deletions(-) -- 1.8.3.1