Re: [PATCH 0/3] enable memcg accounting for kernfs objects

[Roman Gushchin]

On Tue, Aug 09, 2022 at 07:31:31AM -1000, Tejun Heo wrote:
> (cc'ing Johannes)
> 
> Hello,
> 
> On Sun, Jul 31, 2022 at 06:37:15PM +0300, Vasily Averin wrote:
> > 1) creating a new netdevice allocates ~50Kb of memory, where ~10Kb
> >    was allocated for 80+ kernfs nodes.
> > 
> > 2) cgroupv2 mkdir allocates ~60Kb of memory, ~10Kb of them are kernfs
> >    structures.
> > 
> > 3) Shakeel Butt reports that Google has workloads which create 100s
> >    of subcontainers and they have observed high system overhead
> >    without memcg accounting of kernfs.
> 
> So, I don't have anything against accounting kernfs objects in general but,
> for cgroups, because cgroups are what determines what gets charged where,
> I'm not quite sure whether following the usual "charge it to the allocating
> task's cgroup" is the best way to go about it. I wonder whether it'd be
> better to attach it to the new cgroup's nearest ancestor with memcg enabled.

I also like this approach better, however Michal had some arguments against it.
I don't think there is a huge practical difference, so I'm ok with either
approach.

Thanks!