From mboxrd@z Thu Jan 1 00:00:00 1970 From: Avi Kivity Subject: Re: [kvm-devel] [RFC][PATCH 3/4] kvm-vmx: KVM_EXIT_DEBUG on #BP exceptions Date: Sun, 25 May 2008 13:24:27 +0300 Message-ID: <48393E5B.2040904@qumranet.com> References: <482D9198.7040801@web.de> <200805220926.52133.hollisb@us.ibm.com> <48358460.9020805@qumranet.com> <200805221327.43434.hollisb@us.ibm.com> Mime-Version: 1.0 Content-Type: text/plain; charset=UTF-8; format=flowed Content-Transfer-Encoding: 7bit Cc: Jan Kiszka , kvm-devel To: Hollis Blanchard Return-path: Received: from bzq-179-150-194.static.bezeqint.net ([212.179.150.194]:41233 "EHLO il.qumranet.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1762666AbYEYKY3 (ORCPT ); Sun, 25 May 2008 06:24:29 -0400 In-Reply-To: <200805221327.43434.hollisb@us.ibm.com> Sender: kvm-owner@vger.kernel.org List-ID: Hollis Blanchard wrote: > >> We mostly try to keep cpu emulation outside userspace. >> > > Except for this debug stuff? :) > > Except when it's on the wrong side of the complexity/benefit or performance/cleanliness tradeoffs. >> (of course, that depends on what happens on real hardware. Is there a >> machine check pin? or does the cpu generate the exception internally?) >> > > There is a machine check pin, FWIW. It's the chipset (potentially several IO > bridges away) that discovers no device decoded the MMIO address, and > asynchronously notifies the core. > > I'm happy with machine check as an "MMIO failed" flag, so I guess that leaves > us with a separate "inject debug event" ioctl. > In that case it's more reasonable to have a machine check ioctl. Are the other chipset-discovered causes of this? Perhaps on other ppc boards? The x86 analog is triple faults. On real hardware, a triple fault asserts a processor pin and shuts down the processor. The motherboard detects this and causes a system reset. So we have a triple fault exit reason, and reset generation happens in userspace. -- error compiling committee.c: too many arguments to function