On 11/22/2013 05:08 PM, Peter Zijlstra wrote:

On Fri, Nov 22, 2013 at 02:59:31PM +0100, Sebastian Andrzej Siewior wrote:

quoted

+extern void __lockfunc rt_spin_try_unlock(spinlock_t *lock);

I know what you mean, but.. try_unlock() just sounds wrong, how can we
attempt but fail to unlock a lock we hold ;-)

what about
- rt_spin_unlock_try_or_fail();
- __rt_spin_unlock_dont_use_me()

Couldn't we tell lockdep to bend the rules in this case?

/me crawls back under his rock ;-)

:)

Sebastian