From mboxrd@z Thu Jan 1 00:00:00 1970 From: Jamie Lokier Subject: Re: [Qemu-devel] qcow2 corruption observed, fixed by reverting old change Date: Wed, 11 Feb 2009 16:48:14 +0000 Message-ID: <20090211164814.GA7161@shareable.org> References: <20090211070049.GA27821@shareable.org> <4992A108.8070304@suse.de> <20090211114126.GC31997@shareable.org> <4992C77D.4030104@suse.de> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Cc: qemu-devel@nongnu.org, kvm-devel , Laurent Vivier To: Kevin Wolf Return-path: Received: from mail2.shareable.org ([80.68.89.115]:43623 "EHLO mail2.shareable.org" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1755955AbZBKQsT (ORCPT ); Wed, 11 Feb 2009 11:48:19 -0500 Content-Disposition: inline In-Reply-To: <4992C77D.4030104@suse.de> Sender: kvm-owner@vger.kernel.org List-ID: Kevin Wolf wrote: > Besides reviewing the code over and over again, I think the only real > chance is that you can get a non-productive copy of your image and add > some debug code so that we can see at least which code path is causing > problems. I have a copy of my image to reproduce the bug, so I can test patches including diagnostic patches. That's what I did to narrow it down. Being a company mail server, I can't send you the image of course. > > Aside from logic, the code mixes signed 32-bit with unsigned 64-bit > > with unclear naming which would make me nervous. My host is 64-bit, > > by the way. > > I would suspect that simply having a 64 bit host isn't enough to trigger > the problem. These patches were in for half a year now without anyone > noticing such failure. It was just for clarity. If there are any bugs it's more likely to be truncation on a 32 bit host :-) I didn't see any mention of "long", so the code should behave the same on 64-bit and 32-bit hosts. > By the way and completely off-topic: Have you already tried to use the > VHD patches? I would really like to know if they fix your problems. Are those patches in kvm-83? I still have the image that was causing problems way back, and I'm converting it to raw now with kvm-83 to see if it now matches the raw image produced by VPC's own tool. Beyond checking that it reads ok, which it didn't before, I don't know how to test the VPC support properly, but I can try booting the image and see if it at least doesn't fsck^H^H^H^Hscandisk like it used to. I'm not using VPC images any more, we just install Windows into empty QCOW2 or raw images, like everyone else. :-) At some point I may test the VPC support with prebuilt images downloaded from Microsoft - you can too! -- Jamie