Port-xen archive

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Old Index]

Re: Hypervisor Crash, xen4.1



On 05/22/11 20:47, Sean Bruno wrote:
On Thu, 2011-05-19 at 19:58 -0700, Christoph Egger wrote:
On 20.05.11 02:15, Sean Bruno wrote:
Playing around with freebsd non-Xen kernels in hvm mode on the new
hypervisor.  I keep getting some gruesome crashes on it that look like
this:

http://pastebin.com/sc5fjgUA

thoughts?

Please report xen bugs on xen-devel.

Christoph

Well, I've been completely unable to post to this list.  I seem to be
subscribed to it, but I can not figure out what is required to actually
post to it.

So, if someone could please "proxy post" this message for me, I'd
appreciate it.

Sean


You got this answer:

----------------------------------------------------------------------
Do a debug=y build of Xen so that the backtraces are accurate. Also if you know how and the crash in notify_via_xen_event_channel comes up often, worth adding some tracing in there to see which pointer dereference is bad. (Or you could point me at your xen-syms file associated with your new backtraces.)

 -- Keir

----------------------------------------------------------------------


The xen-debug.gz got build with debug=y.

At the top you have this line:

(XEN) RIP: e008:[<ffff82c4801057a9>] notify_via_xen_event_channel+0x5f/0x91

The value in the square brackets may be different with a xen-debug.gz then.

Run a build of your xenkernel41 package but do not run a make clean.
Then beneath the work directory you find the xen-syms file.
There you can do:

gdb xen-syms
(gdb) list *(0x<value in square brackets>)

The output of gdb is what Keir is interested in additionally to the
complete xen kernel crash output.

Christoph


Home | Main Index | Thread Index | Old Index