Messages in this thread | | | Date | Tue, 3 Mar 2009 11:05:04 -0800 (PST) | From | David Rientjes <> | Subject | Re: [patch 2/2] slub: enforce cpuset restrictions for cpu slabs |
| |
On Tue, 3 Mar 2009, Christoph Lameter wrote:
> > There is no alternative solution to prevent egregious amounts of slab to > > be allocated in a disjoint cpuset that is supposedly mem_exclusive. > > The amount of memory is limited by the size of a slab page. If the process > goes beyond that amount then the page allocator will come in and enforce > the boundaries. >
Not until all of the partial slabs on the local node are full, which could be significant given s->min_partial.
This change doesn't affect the fastpath in any significant way for systems that have not configured multiple cpusets; systems that have configured more than one cpuset may have specific NUMA locality requirements that slub is dismissing without this patch.
| |