On Tue, Feb 28, 2012 at 1:48 AM, Peter Zijlstra [off-list ref] wrote:

quoted

BUG: soft lockup - CPU#3 stuck for 22s! [kworker/3:1:78]

So you're stuck in softirq for 22s+, max_restart is 10, this gives that
on average you spend 2.2s+ per softirq invocation, this is completely
absolutely bonkers. Softirq handlers should never consume significant
amount of cpu-time.

I'm not sure whether this was just one __do_softirq() invocation, or
multiple.  I'll see if I can instrument this, but the suspicion is
that max_restart can't help when softirq is constantly re-triggered.
I'll also add the below patch to see if any single invocation is
violating these expectations.

--
Dan