Re: [PATCH] mm/slab: stop exporting kvfree_rcu_barrier[_on_cache]()

Harry Yoo <[email protected]> Mon, 3 Aug 2026 12:26:55 +0000
Newsgroups org.kernel.vger.rcu,org.kernel.vger.linux-kernel,org.kvack.linux-mm
Message-ID <anCCVU7Y6lMaPmAT@dev>
On Mon, Aug 03, 2026 at 09:59:44AM +0200, Vlastimil Babka (SUSE) wrote:
> On 7/30/26 22:59, Paul E. McKenney wrote:
> > On Thu, Jul 30, 2026 at 04:32:46PM +0200, Vlastimil Babka (SUSE) wrote:
> > Which reminds me...  Are you guys interested in taking the kfree_rcu()
> > test code from kernel/rcu/rcuscale.c?  See for example the kfree_rcu_test
> > module parameter.
> 
> Hmm, we have a slub kunit test, but I think no such torture/scaling test
> infra. Maybe once we have that? What do you think, Harry?

That'd be definitely useful. We have been reviving performance tests
every time we wondered if it'd affect performance in a surprising way.
But we've never put that in the mainline ;)

Probably test_kmalloc_nolock_and_friends*() don't even belong
slub_kunit.

Perhaps worth starting small with lib/test_slab.c, and slowly
move/add torture/scaling tests to there?

-- 
Cheers,
Harry / Hyeonggon