Re: sched: long running interrupts breaking spinlocks

[Peter Zijlstra]

On Tue, Apr 08, 2014 at 11:26:56AM -0400, Sasha Levin wrote:
> Hi all,
> 
> (all the below happened inside mm/ code, so while I don't suspect
> it's a mm/ issue you folks got cc'ed anyways!)
> 
> While fuzzing with trinity inside a KVM tools guest running the latest -next
> kernel, I've stumbled on the following:
> 
> [ 4071.166362] BUG: spinlock lockup suspected on CPU#19, trinity-c19/17092

That's a heuristic in the spinlock code; triggering it with big machines
(19 cpus is far bigger than anything at the time that code was written)
and virt (yay for lock owner preemption; another thing we didn't have
back when) is trivial.

I'd not worry too much about this.

So DEBUG_SPINLOCKS turns spin_lock() into something like:

  for (i = 0; i < loops; i++)
  	if (spin_trylock())
		return;

  /* complain */

And you simply ran out of loops.
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/