Re: [Xen-devel] fxsave, fnsave, ltr hang for guest OS.

[Keir Fraser <keir@xxxxxxx>]

On 03/11/2010 01:10, "alarson@xxxxxxxx" <alarson@xxxxxxxx> wrote:

> Thanks Keir, I was headed in that general direction, but I was hoping to
> get lucky.
> 
> I setup the serial cable and did a number of backtraces.  I'm hoping that
> the backtrace for the host is correlated with the client backtrace, but
> I'm suspicious.  Some of the backtraces I got list my client (e.g., on
> CPU1) but don't list a CPU1 host state.  In all cases my clint is doing
> what I expect (attempting to execute an fnsave instruction).
> 
> I slurped the relevant host stacktraces from the console debug log, and I
> don't see a pattern (again hoping to get lucky).  I'll put away my private
> OS shovel, and see if I can start fashioning a Xen shovel.  If you see
> something obvious below that looks suspicious, let me know, otherwise I'll
> see what my Xen shovel digs up.  Thanks again for your help.
> 
> Host CPU traces from the CPU's associated with my OS client.

Do a 'debug=y make' build of Xen and your backtraces will be precise as wthe
build will include frame pointers. Looking at these backtraces, they mostly
seem to be in vmx_intr_assist() doing interrupt delivery work (e.g.,
hvm_isa_irq_assert()). Perhaps you are stuck in some kind of interrupt
delivery loop?

 -- Keir



_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-devel