From mboxrd@z Thu Jan 1 00:00:00 1970 From: Jan Kiszka Subject: [PATCH 0/3] x86: User space NMI injection Date: Mon, 15 Sep 2008 18:28:20 +0200 Message-ID: <48CE8D24.9070007@siemens.com> Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-15 Content-Transfer-Encoding: 7bit Cc: "Yang, Sheng" , Avi Kivity To: kvm-devel Return-path: Received: from lizzard.sbs.de ([194.138.37.39]:22626 "EHLO lizzard.sbs.de" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1753040AbYIORA3 (ORCPT ); Mon, 15 Sep 2008 13:00:29 -0400 Sender: kvm-owner@vger.kernel.org List-ID: This series adds support for injecting NMIs from the QEMU user space into KVM guests. So far NMI injection is limited to the in-kernel APIC emulation. With the patches applied, the 'nmi ' monitor command starts to work as well as NMIs from the user space APIC. The series adds support for VMX, but SVM-related bits should be easy to add. Moreover, I must admit that I was not yet able to test the patches as I post them here because all my Intel boxes lack virtual NMI support - I hacked cpu_has_virtual_nmis to return 1 unconditionally. All typical cases (i.e. non-contended NMI deliveries) should have been covered nevertheless. I'm still playing with ideas and concepts how to work around that lacking NMI support in older (but still very present) Intel CPUs. But whatever the result of this may be, that can also be applied on top of the current implementation. Jan -- Siemens AG, Corporate Technology, CT SE 2 Corporate Competence Center Embedded Linux