Re: [RFC PATCH 0/3] mm, memcg: isolate deprecated v1 state from struct mem_cgroup

[Johannes Weiner]

On Mon, Nov 02, 2026 at 04:52:44PM +0000, Roman Gushchin wrote:
> Tao Cui <cui.tao@xxxxxxxxx> writes:
> 
> > Hi Roman,
> >
> > 在 2026/10/30 14:05, Roman Gushchin 写道:
> >> I believe we want to delete this code sooner than later, why would we rewrite it before that?
> >> 
> >
> > My assumption was that v1 would still be maintained for some time
> > before its eventual removal, and the goal was to reduce the coupling
> > between the remaining v1 code and the shared core structures during
> > that period, rather than to rewrite the implementation for its own
> > sake.
> 
> I understand. But personally I don't see big issues with the current
> code caused by the v1 stuff. Maybe I miss something...
> 
> The proposed changes on the other hand create difficulties for those
> who use v1 and maintain some local changes or simple backport upstream
> fixes. Of course, it's not a reason to not change the upstream code if
> it's justified, but in this case I'm hardly convinced.
> 
> > If the expectation is that the remaining v1 implementation will be
> > removed in the near future, then I agree this work becomes much harder
> > to justify.
> >
> > Am I being too optimistic about how long the remaining v1 code is
> > expected to stay around?
> 
> Well, Idk, it was considered deprecated for quite some time already,
> so maybe announcing that it will be deleted in a year is reasonable?

That might be tougher for v1 users than the member placement :-)

My understanding is that there are still a few heavy v1 users out
there. I'm not sure they can move within that timeframe. Personally I
think the cost of the v1 code now is kind of negligible now?

> Shakeel, Johannes, Michal, wdyt?

I was a weak yes on Tao's patch. Happy to keep it out with you and
Michal not being fans of it.