On Tue 10-03-20 13:41:49, Kirill A. Shutemov wrote:
> On Tue, Mar 10, 2020 at 10:48:36AM +0100, Michal Hocko wrote:
> > [Cc Kirill, I didn't realize he has implemented this code]
>
> My first non-trivial mm contribution :P
Everybody has to pay for sins of youth :p
[...]
> > It seems that the code has been broken since 2c488db27b61 ("memcg: clean
> > up memory thresholds"). We've had 371528caec55 ("mm: memcg: Correct
> > unregistring of events attached to the same eventfd") but it didn't
> > catch this case for some reason. Unless I am missing something the code
> > was broken back then already. Kirill please double check after me.
>
> I think the issue exitsted before 2c488db27b61. The fields had different
> names back then.
>
> The logic to make unregister never-fail is added in 907860ed381a
> ("cgroups: make cftype.unregister_event() void-returning"). I believe the
> Fixes should point there.
Yes, you seem to be right. It doesn't make a difference much as both
went in to the same kernel but a proper Fixes tag is really valuable.
Thanks for looking into that.
--
Michal Hocko
SUSE Labs
