s->cpu_partial determine the maximum number of objects kept
in the per cpu partial lists of a processor. Currently, it is used for
not only per cpu partial list but also cpu freelist. Therefore
get_partial_node() doesn't work properly according to our first intention.
Fix it as forcibly assigning 0 to objects count when we get for cpu freelist.
Signed-off-by: Joonsoo Kim <js1304@xxxxxxxxx>
Cc: Christoph Lameter <cl@xxxxxxxxx>
Cc: David Rientjes <rientjes@xxxxxxxxxx>
diff --git a/mm/slub.c b/mm/slub.c
index efce427..88dca1d 100644
--- a/mm/slub.c
+++ b/mm/slub.c
@@ -1550,7 +1550,12 @@ static void *get_partial_node(struct kmem_cache *s,
c->page = page;
stat(s, ALLOC_FROM_PARTIAL);
object = t;
- available = page->objects - page->inuse;
+
+ /*
+ * We don't want to stop without trying to get
+ * cpu partial slab. So, forcibly set 0 to available
+ */
+ available = 0;
} else {
available = put_cpu_partial(s, page, 0);
stat(s, CPU_PARTIAL_NODE);
--
1.7.9.5
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@xxxxxxxxx. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@xxxxxxxxx";> email@xxxxxxxxx </a>