On (21/06/07 15:54), Maxim Levitsky wrote: > > Overall this looks OK to me. > > Do you have a test case in which this patch helps to make the guest behave better after > the host suspend though? I tested this and I don't see any significant change. > (guest works after host suspend before and after, and I still have clocksource > watchdogs firing in the guest) Yes, we have automated suspend-resume tests. What happens for us is that, when not properly paused, guests, upon host resume, discover jiffies spikes and trigger all sort of watchdogs.
