From mboxrd@z Thu Jan 1 00:00:00 1970 From: Avi Kivity Subject: Re: [PATCH] kvm-userspace: Cleanup user space NMI injection Date: Wed, 26 Nov 2008 18:26:24 +0200 Message-ID: <492D78B0.7080502@redhat.com> References: <492AC80E.8080906@siemens.com> <492D3120.7030902@redhat.com> <492D3674.1070003@siemens.com> Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-15; format=flowed Content-Transfer-Encoding: 7bit Cc: kvm-devel , "Yang, Sheng" To: Jan Kiszka Return-path: Received: from mx2.redhat.com ([66.187.237.31]:42114 "EHLO mx2.redhat.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1754599AbYKZQ0b (ORCPT ); Wed, 26 Nov 2008 11:26:31 -0500 In-Reply-To: <492D3674.1070003@siemens.com> Sender: kvm-owner@vger.kernel.org List-ID: Jan Kiszka wrote: >> Hm. For interrupts we need to check, since the interrupt might be >> deasserted or masked while the window is closed. >> >> Is there no way that this can happen for NMIs? >> > > In our emulation, but I also think in real life, there is no way to > deassert an NMI. Therefore, our user space API just sets the nmi_pending > flag. Masking of NMIs (due to mov ss etc.) is handled by the NMI > injection code, and this has to happen in kernel anyway as the other NMI > sources (APIC and IOAPIC) do not bother about the NMI windows as well. > I think you're right, and applied the patch. Thanks. -- error compiling committee.c: too many arguments to function