RE: [Xen-devel] kernel panic - James Harper - com.xensource.lists.xen-devel

6 messages in com.xensource.lists.xen-develRE: [Xen-devel] kernel panic

From	Sent On	Attachments
James Harper	27 Sep 2004 05:41
Keir Fraser	27 Sep 2004 06:14
James Harper	27 Sep 2004 06:21
Keir Fraser	27 Sep 2004 06:25
Ian Pratt	27 Sep 2004 08:59
James Harper	27 Sep 2004 19:00

Subject:	RE: [Xen-devel] kernel panic
From:	James Harper (Jam...@bendigoit.com.au)
Date:	09/27/2004 07:00:27 PM
List:	com.xensource.lists.xen-devel

Have you got a serial line on the machine? If so, if you have a debug build we should be getting interesting debugging.

Also, before trying to destroy the domain, hit ctrl-A three times and then 'q' and 'o'.

This will do an audit of all the domain's remaining memory, and might give us a clue as to what is broken.

Even more informative might be to add a call to audit_domains() in the "Bad page free for domain" error path in

xen/common/dom_mem_ops.c

Unfortunately I don't have any time to maintain this machine currently - it's in the office and I'm not. So I've disabled wpt and am testing without.

Would it be possible to have xen hook into the syslog facility of dom0, which could then be relayed to another server? Most likely dom0 would be too broken by this point to be able to function in this way though.

What are you doing in the domains that's causing this bug? We've tried to reproduce but can't so far.

The domain that is crashing is a mail server. It's running exim4 + cyrus + mailscanner (+ sendmail + razor + clamav), so it works reasonably hard but is still just doing disk and network io, nothing really special.

Would that error be caused by it running out of memory? I've not seen it even get close to that though, but if a lot of large emails came in in one hit it could happen...

James