From mboxrd@z Thu Jan 1 00:00:00 1970 From: Jeremy Fitzhardinge Subject: Re: i686 vs i586 glibc segfault issue on 64-bit AMD Xen paravirt guests Date: Wed, 02 Sep 2009 10:10:02 -0700 Message-ID: <4A9EA6EA.2040304@goop.org> References: <200909020810.n828AAUe002428@byte-me.mit.edu> Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-15 Content-Transfer-Encoding: 7bit Return-path: In-Reply-To: <200909020810.n828AAUe002428@byte-me.mit.edu> List-Unsubscribe: , List-Post: List-Help: List-Subscribe: , Sender: xen-devel-bounces@lists.xensource.com Errors-To: xen-devel-bounces@lists.xensource.com To: Mitchell E Berger Cc: Christoph Egger , Xen-devel List-Id: xen-devel@lists.xenproject.org On 09/02/09 01:10, Mitchell E Berger wrote: > I apologize for writing to you directly instead of through an officially > supported channel. No problem. > I've filed a bug against glibc in Redhat's Bugzilla > for an issue that only seems to surface on 64-bit Xen paravirt guests > on AMD hosts. Filing this bug with the distro involved seemed to make > sense, but thinking more about it, the bug may truly belong to Xen, > and it seems wrong to cross-post a bug to two completely separate dev > teams' bug trackers at the same time. > I think you're running into a known bug in current mainline Linux kernels, in which 32-bit compat syscalls are not being handled properly on AMD systems. The workaround is to boot with "vdso32=0" on the kernel command line. Unfortunately nobody has taked the time to really work out what's going on (I don't have an AMD system on hand, which makes it awkward). It appears to be a problem with returning from a syscall, rather than the syscall itself. J