Re: [PATCH] mm: vmscan: fix dirty folios throttling on cgroup v1 for MGLRU

[Shakeel Butt]

On Thu, Apr 02, 2026 at 10:51:37AM +0800, Baolin Wang wrote:
> 
> 
[...]
> 
> > > This is v1 only and I don't care much but what is stopping you from moving away
> > > from v1?
> > 
> > For example memsw?
> > 
> > https://lore.kernel.org/linux-mm/q2x4drxpjbxcxgns6bjp446ynsxgl32ckcljqcol7posds4nit@3n3tjq35anvb/
> > 
> > I remember Jingxiang's plan is to improve the page counter first.
> 
> Additionally, we still have some existing users on cgroup v1, and they will
> need some time to migrate to cgroup v2.

Time is fine but is there any specific v1 feature which is a blocker for you.
The memsw is a known one and is the same for Google (data center). I am aiming
to provide memsw like behavior through my memcg_ext proposal but performance is
an open question. I am aiming to have a prototype before LSFMM to further the
discussion.