Re: [PATCH RFC 1/3] slub: keep full slabs on list for per memcg caches

[Vladimir Davydov]

On Wed, May 14, 2014 at 11:16:36AM -0500, Christoph Lameter wrote:
> On Tue, 13 May 2014, Vladimir Davydov wrote:
> 
> > Currently full slabs are only kept on per-node lists for debugging, but
> > we need this feature to reparent per memcg caches, so let's enable it
> > for them too.
> 
> That will significantly impact the fastpaths for alloc and free.
> 
> Also a pretty significant change the logic of the fastpaths since they
> were not designed to handle the full lists. In debug mode all operations
> were only performed by the slow paths and only the slow paths so far
> supported tracking full slabs.

That's the minimal price we have to pay for slab re-parenting, because
w/o it we won't be able to look up for all slabs of a particular per
memcg cache. The question is, can it be tolerated or I'd better try some
other way?

> 
> > @@ -2587,6 +2610,9 @@ static void __slab_free(struct kmem_cache *s, struct page *page,
> >
> >  			} else { /* Needs to be taken off a list */
> >
> > +				if (kmem_cache_has_cpu_partial(s) && !prior)
> > +					new.frozen = 1;
> > +
> >  	                        n = get_node(s, page_to_nid(page));
> 
> Make this code conditional?

No problem, this patch is just a draft. Thanks to static keys, it won't
be difficult to eliminate any overhead if there is no kmem-active
memcgs.

Thanks.

> 
> >  				/*
> >  				 * Speculatively acquire the list_lock.
> > @@ -2606,6 +2632,12 @@ static void __slab_free(struct kmem_cache *s, struct page *page,
> >  		object, new.counters,
> >  		"__slab_free"));
> >
> > +	if (unlikely(n) && new.frozen && !was_frozen) {
> > +		remove_full(s, n, page);
> > +		spin_unlock_irqrestore(&n->list_lock, flags);
> > +		n = NULL;
> > +	}
> > +
> >  	if (likely(!n)) {
> 
> Here too.
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/