Re: [PATCH v3] cgroup: avoid flushing global workqueue in cgroup1_pidlist_destroy_all

[Junnan Zhang]

Hi Tejun,

Fair enough -- the lab numbers demonstrate the coupling but don't
bridge the gap to the 120s we saw in the field, and I can't rule out
that other factors dominated there. Without the vmcore from the
customer I have no way to settle that part.

If we manage to reproduce the hung task splat itself in the lab, or
get usable field data, I'll come back with it.

Thanks to you, Michal and Ridong for the time spent reviewing this.

Junnan