On Thu 02-02-23 10:24:08, Shakeel Butt wrote:
> On Thu, Feb 2, 2023 at 12:27 AM Michal Hocko <mhocko@xxxxxxxx> wrote:
> >
> > On Thu 02-02-23 12:32:57, zhaoyang.huang wrote:
> > > From: Zhaoyang Huang <zhaoyang.huang@xxxxxxxxxx>
> > >
> > > Introducing memory.root_current for the memory charges on root_mem_cgroup.
> >
> > Charges are not currently accounted for the root memcg universally. See
> > try_charge which is used for all user space and skmem charges. I am not
> > 100% sure about objcg based accounting because there is no explicit
> > check for the root memcg but this might be hidden somewhere as well.
>
> Yes in __get_obj_cgroup_from_memcg(). However the reason to use
> try_charge_memcg() to bypass root check was to avoid the race with
> reparenting. More details in c5c8b16b596e ("mm: memcontrol: fix
> root_mem_cgroup charging").
Thanks for the pointer!
> > That means that the patch as is doesn't really provide and usable value.
> > The root exemption has been removed in the past but that has been
> > reverted due to a regression. See ce00a967377b ("mm: memcontrol: revert
> > use of root_mem_cgroup res_counter") for more.
> >
>
> One advantage I can see is if someone is looking for usage for all top
> containers (alive or zombie) but I wanted to know if that was the real
> motivation behind the patch.
Isn't that just a global stats that we already display via /proc files?
--
Michal Hocko
SUSE Labs
