[PATCH -v5 00/14] the saga of FUTEX_UNLOCK_PI wobbles continues

[Peter Zijlstra]

Hi all,

Here be the 5th iteration of these patches; which strive to unlock the rt_mutex
without holding hb->lock.

In total this patch-set is very much like the previous one (including the 'late'
fix) but it gets there in a slightly different -- and hopefully easier to
understand -- route.

The last 3 patches could be superfluous if we can proof that the -EAGAIN retry
in futex_unlock_pi() isn't a problem for bounded execution (or are willing to
accept it). In this case the patch-set ends up simpler than before.

I sincerely hope this is the last of it; but please consider carefully, there be
dragons here.