Currently, slab_mem_going_going_callback() checks whether the node has
N_NORMAL memory in order to be set in slab_nodes.
While it is true that gettind rid of that enforcing would mean
ending up with movables nodes in slab_nodes, the memory waste that comes
with that is negligible.

So stop checking for status_change_nid_normal and just use status_change_nid
instead which works for both types of memory.

Also, once we allocate the kmem_cache_node cache  for the node in
slab_mem_online_callback(), we never deallocate it in
slab_mem_off_callback() when the node goes memoryless, so we can just
get rid of it.

The only side effect is that we will stop clearing the node from slab_nodes.

Suggested-by: David Hildenbrand <david@redhat.com>
Signed-off-by: Oscar Salvador <osalvador@suse.de>
---
 mm/slub.c | 34 +++-------------------------------
 1 file changed, 3 insertions(+), 31 deletions(-)

diff --git a/mm/slub.c b/mm/slub.c
index dc9e729e1d26..236d679e6a1d 100644
--- a/mm/slub.c
+++ b/mm/slub.c
@@ -447,7 +447,7 @@ static inline struct kmem_cache_node *get_node(struct kmem_cache *s, int node)
 
 /*
  * Tracks for which NUMA nodes we have kmem_cache_nodes allocated.
- * Corresponds to node_state[N_NORMAL_MEMORY], but can temporarily
+ * Corresponds to node_state[N_MEMORY], but can temporarily
  * differ during memory hotplug/hotremove operations.
  * Protected by slab_mutex.
  */
@@ -6174,36 +6174,12 @@ static int slab_mem_going_offline_callback(void *arg)
 	return 0;
 }
 
-static void slab_mem_offline_callback(void *arg)
-{
-	struct memory_notify *marg = arg;
-	int offline_node;
-
-	offline_node = marg->status_change_nid_normal;
-
-	/*
-	 * If the node still has available memory. we need kmem_cache_node
-	 * for it yet.
-	 */
-	if (offline_node < 0)
-		return;
-
-	mutex_lock(&slab_mutex);
-	node_clear(offline_node, slab_nodes);
-	/*
-	 * We no longer free kmem_cache_node structures here, as it would be
-	 * racy with all get_node() users, and infeasible to protect them with
-	 * slab_mutex.
-	 */
-	mutex_unlock(&slab_mutex);
-}
-
 static int slab_mem_going_online_callback(void *arg)
 {
 	struct kmem_cache_node *n;
 	struct kmem_cache *s;
 	struct memory_notify *marg = arg;
-	int nid = marg->status_change_nid_normal;
+	int nid = marg->status_change_nid;
 	int ret = 0;
 
 	/*
@@ -6261,10 +6237,6 @@ static int slab_memory_callback(struct notifier_block *self,
 	case MEM_GOING_OFFLINE:
 		ret = slab_mem_going_offline_callback(arg);
 		break;
-	case MEM_OFFLINE:
-	case MEM_CANCEL_ONLINE:
-		slab_mem_offline_callback(arg);
-		break;
 	case MEM_ONLINE:
 	case MEM_CANCEL_OFFLINE:
 		break;
@@ -6335,7 +6307,7 @@ void __init kmem_cache_init(void)
 	 * Initialize the nodemask for which we will allocate per node
 	 * structures. Here we don't need taking slab_mutex yet.
 	 */
-	for_each_node_state(node, N_NORMAL_MEMORY)
+	for_each_node_state(node, N_MEMORY)
 		node_set(node, slab_nodes);
 
 	create_boot_cache(kmem_cache_node, "kmem_cache_node",
-- 
2.49.0

On 5/2/25 10:36, Oscar Salvador wrote:
> Currently, slab_mem_going_going_callback() checks whether the node has
> N_NORMAL memory in order to be set in slab_nodes.
> While it is true that gettind rid of that enforcing would mean
> ending up with movables nodes in slab_nodes, the memory waste that comes
> with that is negligible.
> 
> So stop checking for status_change_nid_normal and just use status_change_nid
> instead which works for both types of memory.
> 
> Also, once we allocate the kmem_cache_node cache  for the node in

					     ^ object? instance?

> slab_mem_online_callback(), we never deallocate it in
> slab_mem_off_callback() when the node goes memoryless, so we can just
> get rid of it.
> 
> The only side effect is that we will stop clearing the node from slab_nodes.

Technically, another effect is that also any newly created kmem caches after
 node hotremove will now allocate their kmem_cache_node for the node(s) that
was hotremoved. But that should be even more negligible waste than from
using N_MEMORY.

> Suggested-by: David Hildenbrand <david@redhat.com>
> Signed-off-by: Oscar Salvador <osalvador@suse.de>

Reviewed-by: Vlastimil Babka <vbabka@suse.cz>