On Mon, 16 Mar 2015 14:59:10 +0100
Sebastian Andrzej Siewior [off-list ref] wrote:

* Steven Rostedt | 2015-03-12 15:13:07 [-0400]:

quoted

Please scream at me if I messed something up. Please test the patches too.

So Paul remided us about the dead lock thingy that has been reported.
Users reported that it does not occur with v3.18-RT and they think it is
due to 'Revert "timers: do not raise softirq unconditionally"' in 
Revert-timers-do-not-raise-softirq-unconditionally.patch.

I reverted it because I couldn't get highres to get to work at all on
v3.18 due to different synchronisation / expectaion of the timer
framework. Since the trylock might record a different lock owner it is
possible that this causes the deadlock (it thinks). Therefore it has no
stable tag nor any reference to the deadlock problem.

I guess the question is, is there any other place that does a trylock
in hard irq context? If so, the revert isn't going to fix it.

-- Steve

With this patch applied FULL_NOHZ should not properly work (again) due
to timer softirq wake ups (but this is a different problem).

Sebastian