Re: [Xen-devel] HVM SMP linux guest hangs in cpu_idle() with EFLAGS.IF = 1

[Gianni Tedesco <gianni.tedesco@xxxxxxxxxx>]

On Thu, 2010-07-08 at 14:28 +0100, George Dunlap wrote:
> On Thu, Jul 8, 2010 at 12:55 PM, Gianni Tedesco
> <gianni.tedesco@xxxxxxxxxx> wrote:
> > Hmm, yeah, usually that's a headache to do for one device never mind the
> > whole system...
> 
> But realistically, there's only a handful of devices which it might be
> waiting on -- seems like the disk is the most likely culprit.
> 
> I'm happy to help with the tracing / analysis bit.
> 
> In any case, thanks for doing this, and good luck.

Problem is I seem to have ruled out most of that now. IDE is firing off
IRQ's and the host ACKing them properly. There are even hangs during
periods of no IDE activity - just working from ramdisks. Timers are
getting through. Networking always seems to work fine and the hang
occurs regardless of e1000 vs rtl8139. The bug reproduces without serial
and regardless of acpi, std-vga, etc so there isn't much else that could
be going wrong here.

I managed to get a shell up and running while the system is hung and
just spawning "busybox ls" also hangs in an uninterruptible state
(although the uninterruptable part may have been due to lack of job
control in the shell.)

Since the device model and IRQ delivery seems to work as expected I am
wondering if there could be an artefact that causes, for example, the
kernels semaphore implementation to not work as it should?

Gianni


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