net/mlx4_en: allocate non 0-order pages for RX ring with __GFP_NOMEMALLOC
authorKonstantin Khlebnikov <khlebnikov@yandex-team.ru>
Mon, 18 Apr 2016 11:33:54 +0000 (14:33 +0300)
committerDavid S. Miller <davem@davemloft.net>
Wed, 20 Apr 2016 00:04:24 +0000 (20:04 -0400)
High order pages are optional here since commit 51151a16a60f ("mlx4: allow
order-0 memory allocations in RX path"), so here is no reason for depleting
reserves. Generic __netdev_alloc_frag() implements the same logic.

Signed-off-by: Konstantin Khlebnikov <khlebnikov@yandex-team.ru>
Acked-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
drivers/net/ethernet/mellanox/mlx4/en_rx.c

index 86bcfe510e4e5b930b94de462f5ff9a76d4183c4..56d070d9cf015228075bd06b18769ca3b624083d 100644 (file)
@@ -61,7 +61,7 @@ static int mlx4_alloc_pages(struct mlx4_en_priv *priv,
                gfp_t gfp = _gfp;
 
                if (order)
-                       gfp |= __GFP_COMP | __GFP_NOWARN;
+                       gfp |= __GFP_COMP | __GFP_NOWARN | __GFP_NOMEMALLOC;
                page = alloc_pages(gfp, order);
                if (likely(page))
                        break;