On Thu, Oct 24, 2024 at 06:54:01PM GMT, Roman Gushchin wrote:
> On Thu, Oct 24, 2024 at 10:23:49AM -0700, Shakeel Butt wrote:
> > On Thu, Oct 24, 2024 at 11:16:52AM GMT, Michal Hocko wrote:
> > > On Wed 23-10-24 23:57:12, 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@xxxxxxxxx>
> > > > ---
> > > > -/**
> > > > - * folio_memcg_lock - Bind a folio to its memcg.
> > > > - * @folio: The folio.
> > > > - *
> > > > - * This function prevents unlocked LRU folios from being moved to
> > > > - * another cgroup.
> > > > - *
> > > > - * It ensures lifetime of the bound memcg. The caller is responsible
> > > > - * for the lifetime of the folio.
> > > > - */
> > > > -void folio_memcg_lock(struct folio *folio)
> > > > -{
> > > > - struct mem_cgroup *memcg;
> > > > - unsigned long flags;
> > > > -
> > > > - /*
> > > > - * The RCU lock is held throughout the transaction. The fast
> > > > - * path can get away without acquiring the memcg->move_lock
> > > > - * because page moving starts with an RCU grace period.
> > > > - */
> > > > - rcu_read_lock();
> > >
> > > Is it safe to remove the implicit RCU?
> >
> > Good question. I think it will be safe to keep the RCU in this patch and
> > in the followup examine each place and decide to remove RCU or not.
>
> I took a really quick look and based on it I believe it is safe.
> Shakeel, can you, please, check too and preferably keep your code intact.
> I think it's better to remove it all together, rather than do it in two steps.
> If we really need rcu somewhere, we can replace folio_memcg_lock()/unlock()
> with an explicit rcu_read_lock()/rcu_read_unlock().
>
Yup going through that and till now it seems safe. Hopefully I will have
the update by the evening.
