Re: [PATCH 4/5] mm: rework non-root kmem_cache lifecycle management

[Roman Gushchin <guro@fb.com>]

On Thu, Apr 18, 2019 at 01:38:44PM +0000, Christopher Lameter wrote:
> On Wed, 17 Apr 2019, Roman Gushchin wrote:
> 
> >  static __always_inline int memcg_charge_slab(struct page *page,
> >  					     gfp_t gfp, int order,
> >  					     struct kmem_cache *s)
> >  {
> > -	if (is_root_cache(s))
> > +	int idx = (s->flags & SLAB_RECLAIM_ACCOUNT) ?
> > +		NR_SLAB_RECLAIMABLE : NR_SLAB_UNRECLAIMABLE;
> > +	struct mem_cgroup *memcg;
> > +	struct lruvec *lruvec;
> > +	int ret;
> > +
> > +	if (is_root_cache(s)) {
> > +		mod_node_page_state(page_pgdat(page), idx, 1 << order);
> 
> Hmmm... This is functionality that is not memcg specific being moved into
> a memcg function??? Maybe rename the function to indicate that it is not
> memcg specific and add the proper #ifdefs?
> 
> >  static __always_inline void memcg_uncharge_slab(struct page *page, int order,
> >  						struct kmem_cache *s)
> >  {
> > -	memcg_kmem_uncharge(page, order);
> > +	int idx = (s->flags & SLAB_RECLAIM_ACCOUNT) ?
> > +		NR_SLAB_RECLAIMABLE : NR_SLAB_UNRECLAIMABLE;
> > +	struct mem_cgroup *memcg;
> > +	struct lruvec *lruvec;
> > +
> > +	if (is_root_cache(s)) {
> > +		mod_node_page_state(page_pgdat(page), idx, -(1 << order));
> > +		return;
> > +	}
> 
> And again.
> 

Good point! Will do in v2.

Thanks!