Re: perf/tracepoint: another fuzzer generated lockup

[Frederic Weisbecker <fweisbec@gmail.com>]

On Sat, Nov 09, 2013 at 04:13:56PM +0100, Peter Zijlstra wrote:
> On Sat, Nov 09, 2013 at 03:52:59PM +0100, Frederic Weisbecker wrote:
> > So, an idea of what may be happening: an event overflows while FASYNC flag is set so it triggers an irq work
> > to send the signal (kill_fasync).
> > After the irq work triggers, it generates an irq_work_exit event, which in turn overflows and,
> > if it has FASYNC, triggers a new irq work. The irq work triggers and generates an irq work exit event which
> > has FASYNC flag, etc...
> > 
> > Looks like a nice way to deadlock with an infinite loop of irq work.
> 
> 
> Yep, exactly, see the email I just send.

In fact, raising an irq work from an irq work should simply be prohibited. That's not a sane
behaviour.

It's natural for async stuffs that have reasonable delays between each pass allow re-enqueuing,
like workqueue or rcu callbacks, or timers. But with irq work that doesn't look right, expect for
lazy irq works though. But lets just not allow it at all :)