Re: [PATCH 3/6] sched/numa: fix scan-period adjustment for shared faults

[Gregory Price]

On Thu, Oct 15, 2026 at 07:39:42PM +0800, Li Zhe wrote:
> So I think I should separate the severe fixes from the heuristic changes.

This would be appreciated.  The heuristic changes that aren't obvious
Fixes: should not be lumped in.

> For the latter, I will narrow the next version to NUMA balancing behavior
> that is specific to memory tiering, instead of trying to justify the
> changes as general improvements for normal NUMA balancing workloads.

For heuristic changes, please try to generate some kind of data that
demonstrate the behavior being changed. Some of this scan-period tweaking
is very difficult to understand how it will affect end-users without
examples.

There's risk of introducing unexpected regressions when we change these
heuristics.  For example - the recent shmem scanning fix.  Fixing that
resulted in 4-5 other regressions, despite being a "correctness" change.

~Gregory