On Sun, Mar 8, 2020, 08:02 kernel test robot <rong.a.chen@intel.com> wrote:

>
> FYI, we noticed a -97.8% regression of will-it-scale.per_process_ops due
> to commit:
>
> commit: 8019ad13ef7f64be44d4f892af9c840179009254 ("futex: Fix inode
> life-time issue")


Well, that's not optimal.

Peter, any ideas? One of the things that worried me about changing to an
u64 was what I think it causes extra padding, I think. And maybe that
messes with the comparison and hashing of the futexes?