From mboxrd@z Thu Jan 1 00:00:00 1970 From: Jan Kiszka Subject: Re: Booting up an old Windows 3.1 harddisk image Date: Tue, 18 Oct 2011 15:21:49 +0200 Message-ID: <4E9D7D6D.8050308@siemens.com> References: <20124.53132.479840.952606@quad.stoffel.home> <4E9D57FE.6050903@redhat.com> <4E9D7A17.9060107@siemens.com> <4E9D7B83.4020402@redhat.com> Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: QUOTED-PRINTABLE Cc: John Stoffel , "kvm@vger.kernel.org" To: Avi Kivity Return-path: Received: from thoth.sbs.de ([192.35.17.2]:15822 "EHLO thoth.sbs.de" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1757866Ab1JRNWE (ORCPT ); Tue, 18 Oct 2011 09:22:04 -0400 In-Reply-To: <4E9D7B83.4020402@redhat.com> Sender: kvm-owner@vger.kernel.org List-ID: On 2011-10-18 15:13, Avi Kivity wrote: > On 10/18/2011 03:07 PM, Jan Kiszka wrote: >>> >>> It's probably a bug in kvm. Try collecting a complete trace as in >>> http://www.linux-kvm.org/page/Tracing and posting it here, maybe >>> something will jump out. >>> >>> Also try out -no-kvm. >> >> A shot from the hips: J=F6rg and I are just hunting down 16 bit task >> switch issues on AMD hosts. If this problem may be related can be >> checked by loading kvm-amd with npt=3D0, ie. disabling nested paging= =2E >=20 > Do you mean that you suspect that kvm's task switch emulation is more > accurate than the hardware? Gleb will have a heart attack. As it looks like (for AMD). But the last proof is yet missing - /me is instrumenting right now. > Another area to look at is A20 emulation, based on where the hang occ= urs. Yeah, maybe. Traces may tell. Jan --=20 Siemens AG, Corporate Technology, CT T DE IT 1 Corporate Competence Center Embedded Linux