Displaying 1 result from an estimated 1 matches for "poll_irq".
Did you mean:
poll_in
2012 Oct 17
28
Xen PVM: Strange lockups when running PostgreSQL load
...e for cpu#0 and
cpu#7 (doing my tests with 8 VCPUs). So apparently both cpus were
waiting on the slow path for it to become free. Though actually
it was free! Now, here is one issue I have in understanding the
dump: the back traces produced in crash are in the normal form
not showing any cpu in the poll_irq HV call. Only when using
the form that uses the last known stack location and displays
all text symols found will get close for cpu#7. cpu#0 still does
not seem to be anywhere close. This could be a problem with crash,
or with the way PVM works, I am not sure.
Anyway, from what I could take from t...