On Sat, Aug 09, 2025 at 08:32:22AM -1000, Tejun Heo <tj@kernel.org> wrote:
> Also, as Shakeel already pointed out, this would need to be accumulated
> hierarchically. The tricky thing is determining how the accumulation should
> work. Hierarchical summing up is simple and we can use the usual rstat
> propagation; however, that would deviate from how pressure durations are
> propagated for .pressure metrics, where each cgroup tracks all / some
> contention states in its descendants. For simplicity's sake and if the
> number ends up in memory.stat, I think simple summing up should be fine as
> long as it's so noted in the documentation. Note that this semantical
> difference would be another reason to avoid the "pressure" name.

One more point to clarify -- should the value include throttling from
ancestors or not. (I think both are fine but) this semantic should also
be described in the docs. I.e. current proposal is
	value = sum_children + self
and if you're see that C's value is 0, it doesn't mean its sockets
weren't subject of throttling. It just means you need to check also
values in C ancestors. Does that work?

Thanks,
Michal