On Wed, Nov 18, 2020 at 12:25:38PM +0100, Vlastimil Babka wrote:
> On 11/18/20 9:27 AM, Bharata B Rao wrote:
> > The page order of the slab that gets chosen for a given slab
> > cache depends on the number of objects that can be fit in the
> > slab while meeting other requirements. We start with a value
> > of minimum objects based on nr_cpu_ids that is driven by
> > possible number of CPUs and hence could be higher than the
> > actual number of CPUs present in the system. This leads to
> > calculate_order() chosing a page order that is on the higher
> > side leading to increased slab memory consumption on systems
> > that have bigger page sizes.
> >
> > Hence rely on the number of online CPUs when determining the
> > mininum objects, thereby increasing the chances of chosing
> > a lower conservative page order for the slab.
> >
> > Signed-off-by: Bharata B Rao <bharata@xxxxxxxxxxxxx>
>
> Acked-by: Vlastimil Babka <vbabka@xxxxxxx>
>
> Ideally, we would react to hotplug events and update existing caches
> accordingly. But for that, recalculation of order for existing caches would
> have to be made safe, while not affecting hot paths. We have removed the
> sysfs interface with 32a6f409b693 ("mm, slub: remove runtime allocation
> order changes") as it didn't seem easy and worth the trouble.
>
> In case somebody wants to start with a large order right from the boot
> because they know they will hotplug lots of cpus later, they can use
> slub_min_objects= boot param to override this heuristic. So in case this
> change regresses somebody's performance, there's a way around it and thus
> the risk is low IMHO.
I agree. For the absolute majority of users there will be no difference.
And there is a good workaround for the rest.
Acked-by: Roman Gushchin <guro@xxxxxx>
