Re: [RFC PATCH 3/3] memcg-v1: remove memcg move locking code

[Michal Hocko <mhocko@suse.com>]

On Thu 24-10-24 13:32:53, Yosry Ahmed wrote:
> On Thu, Oct 24, 2024 at 12:45 PM Michal Hocko <mhocko@suse.com> wrote:
> >
> > On Thu 24-10-24 10:26:15, Shakeel Butt wrote:
> > > On Thu, Oct 24, 2024 at 04:50:37PM GMT, Roman Gushchin wrote:
> > > > On Wed, Oct 23, 2024 at 11:57:12PM -0700, Shakeel Butt wrote:
> > > > > The memcg v1's charge move feature has been deprecated. There is no need
> > > > > to have any locking or protection against the moving charge. Let's
> > > > > proceed to remove all the locking code related to charge moving.
> > > > >
> > > > > Signed-off-by: Shakeel Butt <shakeel.butt@linux.dev>
> > > >
> > > > Reviewed-by: Roman Gushchin <roman.gushchin@linux.dev>
> > >
> > > Thanks Roman for the review. Based on Michal's question, I am planning
> > > to keep the RCU locking in the next version of this patch and folowup
> > > with clear understanding where we really need RCU and where we don't.
> >
> > I think it would be safer and easier to review if we drop each RCU
> > separately or in smaller batches.
> 
> FWIW if we go with this route, I agree with Roman's idea about
> replacing folio_memcg_lock()/unlock()
> with an explicit rcu_read_lock()/rcu_read_unlock(), and then having
> separate patches/series that remove the RCU annotations. If done in a
> separate series, we should comment the explicit RCU calls
> appropriately to reflect the fact that they should mostly be removed
> (or at least re-evaluated).

Agreed!

-- 
Michal Hocko
SUSE Labs