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

[Junnan Zhang]

Hi Tejun,

Thanks for the clarification -- agreed that the 120s stall needs to
be convincingly attributed before this can count as a minimal fix.

I'll put together a reproducer that drives only cgroup churn plus
cgroup.procs readers on an otherwise idle machine, with measured
cgroup1_pidlist_destroy_all() latencies (and a lowered
hung_task_timeout_secs to catch actual splats), and report back with
the numbers. It'll take me a few days to set up.

Thanks,
Junnan