Re: [External] Re: [PATCH 4/5] mm: memcg/slab: Fix root memcg vmstats

[Roman Gushchin]

On Thu, Oct 29, 2020 at 02:15:43PM +0800, Muchun Song wrote:
> On Thu, Oct 29, 2020 at 8:14 AM Roman Gushchin <guro@xxxxxx> wrote:
> >
> > On Wed, Oct 28, 2020 at 10:56:20AM +0800, Muchun Song wrote:
> > > On Wed, Oct 28, 2020 at 2:48 AM Roman Gushchin <guro@xxxxxx> wrote:
> > > >
> > > > On Tue, Oct 27, 2020 at 04:02:55PM +0800, Muchun Song wrote:
> > > > > If we reparent the slab objects to the root memcg, when we free
> > > > > the slab object, we need to update the per-memcg vmstats to keep
> > > > > it correct for the root memcg. Now this at least affects the vmstat
> > > > > of NR_KERNEL_STACK_KB for !CONFIG_VMAP_STACK when the thread stack
> > > > > size is smaller than the PAGE_SIZE.
> > > > >
> > > > > Fixes: ec9f02384f60 ("mm: workingset: fix vmstat counters for shadow nodes")
> > > > > Signed-off-by: Muchun Song <songmuchun@xxxxxxxxxxxxx>
> > > >
> > > > Can you, please, drop this patch for now?
> > > >
> > > > I'm working on a bigger cleanup related to the handling of the root memory
> > > > cgroup (I sent a link earlier in this thread), which already does a similar change.
> > > > There are several issues like this one, so it will be nice to fix them all at once.
> > >
> > > I have read the patch of https://lkml.org/lkml/2020/10/14/869. You
> > > mean this patch
> > > fixes this issue? It chooses to uncharge the root memcg. But here we may need to
> > > uncharge the root memcg to keep root vmstats correct. If we do not do
> > > this, we can
> > > see the wrong vmstats via root memory.stat(e.g. NR_KERNEL_STACK_KB).
> >
> > I pointed at a different patch in the same thread (it looks like you read the first one):
> > https://lkml.org/lkml/2020/10/21/612

Hi Muchun!

Can you please, resend your patch? The planned cleanup of the root memory cgroup
is more complex than expected, so I think it makes sense to merge your patch without
waiting for it. I'm sorry for delaying it initially.

Please, feel free to add
Acked-by: Roman Gushchin <guro@xxxxxx>

Thank you!

Roman