Subject: Re: x86 changes
To: Manuel Bouyer <bouyer@antioche.eu.org>
From: Andrew Doran <ad@NetBSD.org>
List: port-xen
Date: 09/27/2007 23:24:46
On Thu, Sep 27, 2007 at 10:41:09PM +0200, Manuel Bouyer wrote:
> On Thu, Sep 27, 2007 at 08:55:47PM +0200, Manuel Bouyer wrote:
> > On Thu, Sep 27, 2007 at 08:42:58PM +0200, Manuel Bouyer wrote:
> > > [...]
> > > mutex_enter(0,bbafc910,c6561e00,4,c7e54c40) at netbsd:mutex_enter+0xd
> > > cpu_switchto(c7e54c40,c0d6ac50,c840f8dc,c0482518,0) at netbsd:cpu_switchto+0x81
> >
> > netbsd:cpu_switchto+0x75: movl 0x18(%edi),%ecx
> > db>
> > netbsd:cpu_switchto+0x78: pushl 0x34(%ecx)
> > db>
> > netbsd:cpu_switchto+0x7b: pushl %eax
> > db>
> > netbsd:cpu_switchto+0x7c: call netbsd:ras_lookup
> > db>
> > netbsd:cpu_switchto+0x81: addl $0x8,%esp
> > db>
> > netbsd:cpu_switchto+0x84: cmpl $-0x1,%eax
> > db>
> > netbsd:cpu_switchto+0x87: jz 0xc0100147
> >
> > The mutex_enter could be called from ras_lookup(), but I'm not sure why
> > ras_lookup doesn't appear in the stack trace ...
>
> This seems to be related to threads. non-threaded apps doesn't seem to
> trigger the issue, but just starting python2.4 is enough to get an instant
> panic.
>
> Would there be some other place in the xen machep files where we'd need to
> set the curlwp now ?
I think it's an issue in the RAS check in cpu_switchto. I'll have a look
tomorrow morning.
Thanks,
Andrew