Re: [PATCH v2 2/7] memcg: dynamically allocate lruvec_stats

[Shakeel Butt <shakeel.butt@linux.dev>]

On Mon, Apr 29, 2024 at 08:50:11AM -0700, Roman Gushchin wrote:
> On Fri, Apr 26, 2024 at 05:37:28PM -0700, Shakeel Butt wrote:
[...]
> > +unsigned long lruvec_page_state_local(struct lruvec *lruvec,
> > +				      enum node_stat_item idx)
> > +{
> > +	struct mem_cgroup_per_node *pn;
> > +	long x = 0;
> > +
> > +	if (mem_cgroup_disabled())
> > +		return node_page_state(lruvec_pgdat(lruvec), idx);
> > +
> > +	pn = container_of(lruvec, struct mem_cgroup_per_node, lruvec);
> > +	x = READ_ONCE(pn->lruvec_stats->state_local[idx]);
> > +#ifdef CONFIG_SMP
> > +	if (x < 0)
> > +		x = 0;
> > +#endif
> 
> Not directly related to your change, but do we still need it? And if yes,
> do we really care about !CONFIG_SMP case enough to justify these #ifdefs?
> 

That's a good question and I think this is still needed. Particularly on
large machines with large number of CPUs, we can have a situation where
the flusher is flushing the CPU 100 and in parallel some workload
allocated a lot of pages on, let's say, CPU 0 and freed on CPU 200.

> > +	return x;
> > +}
> > +
> >  /* Subset of vm_event_item to report for memcg event stats */
> >  static const unsigned int memcg_vm_event_stat[] = {
> >  	PGPGIN,
> > @@ -5492,18 +5546,25 @@ static int alloc_mem_cgroup_per_node_info(struct mem_cgroup *memcg, int node)
> >  	if (!pn)
> >  		return 1;
> >  
> > +	pn->lruvec_stats = kzalloc_node(sizeof(struct lruvec_stats), GFP_KERNEL,
> > +					node);
> 
> Why not GFP_KERNEL_ACCOUNT?
> 

Previously struct lruvec_stats was part of struct mem_cgroup_per_node
and we use GFP_KERNEL to allocate struct mem_cgroup_per_node. I kept the
behavior same and if we want to switch to GFP_KERNEL_ACCOUNT, I think it
should be a separate patch.

Thanks for the review.
Shakeel