[PATCH] mm/slub: let number of present CPUs determine the slub page order

[Bharata B Rao]

Commit 045ab8c9487b ("mm/slub: let number of online CPUs determine
the slub page order") changed the slub page order to depend on
num_online_cpus() from nr_cpu_ids. However we find that certain
caches (kmalloc) are initialized even before the secondary CPUs
are onlined resulting in lower slub page order and subsequent
regression.

Switch to num_present_cpus() instead.

Reported-by: Vincent Guittot <vincent.guittot@xxxxxxxxxx>
Signed-off-by: Bharata B Rao <bharata@xxxxxxxxxxxxx>
Fixes: 045ab8c9487b ("mm/slub: let number of online CPUs determine the slub page order")
---
 mm/slub.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/mm/slub.c b/mm/slub.c
index d9e4e10683cc..2f3e412c849d 100644
--- a/mm/slub.c
+++ b/mm/slub.c
@@ -3433,7 +3433,7 @@ static inline int calculate_order(unsigned int size)
 	 */
 	min_objects = slub_min_objects;
 	if (!min_objects)
-		min_objects = 4 * (fls(num_online_cpus()) + 1);
+		min_objects = 4 * (fls(num_present_cpus()) + 1);
 	max_objects = order_objects(slub_max_order, size);
 	min_objects = min(min_objects, max_objects);
 
-- 
2.26.2