Re: [Xen-devel] Re: [patch] xenfb: fix xenfb suspend/resume race

On 01/04/11 00:34, Konrad Rzeszutek Wilk wrote:

quoted

quoted

I am unclear from your description whether the patch fixes
the problem (I would presume so). Or does it take a long time
to hit this race?

Yes, more than 100 migrations. we hit this issue around 3 times.

OK, so you are still trying to find the culprit.

Did you look at this patch from Ian:

https://patchwork.kernel.org/patch/403192/

We have reproduced the issue with the patch.

?

quoted

I dumped vmcore when guest crashed, from vmcore everything
looked good, fb_info, xenfb_info and so on.

And the event channels are correct?

.. snip..

quoted

diff --git a/drivers/xen/events.c b/drivers/xen/events.c
index ac7b42f..4cfb5e2 100644
--- a/drivers/xen/events.c
+++ b/drivers/xen/events.c

@@ -175,6 +175,8 @@ static struct irq_info *info_for_irq(unsigned irq)
 
 static unsigned int evtchn_from_irq(unsigned irq)
 {
+	if (unlikely(irq < 0 || irq >= nr_irqs))
+		return 0;

You could insert a WARN_ON here to see see if you get this during your
migration process.

Or use xen_raw_printk in case the guest is hung for good.

Thanks for your advice, will try it.