No numbers to back this up, but it seemed obvious to me, that if there
are competing lru_add_drain_all()ers, the work will be minimized if each
flushes its own local queues before locking and doing cross-CPU drains.

Signed-off-by: Hugh Dickins <hughd@google.com>
---
 mm/swap.c | 3 +++
 1 file changed, 3 insertions(+)

diff --git a/mm/swap.c b/mm/swap.c
index 17438fd1f51a..4951f4d35e40 100644
--- a/mm/swap.c
+++ b/mm/swap.c
@@ -834,6 +834,9 @@ static inline void __lru_add_drain_all(bool force_all_cpus)
 	 */
 	this_gen = smp_load_acquire(&lru_drain_gen);
 
+	/* It helps everyone if we do our own local drain immediately. */
+	lru_add_drain();
+
 	mutex_lock(&lock);
 
 	/*
-- 
2.51.0

On 31.08.25 11:18, Hugh Dickins wrote:
> No numbers to back this up, but it seemed obvious to me, that if there
> are competing lru_add_drain_all()ers, the work will be minimized if each
> flushes its own local queues before locking and doing cross-CPU drains.
> 

Sounds reasonable

Acked-by: David Hildenbrand <david@redhat.com>

-- 
Cheers

David / dhildenb