[net-next,3/3] net: Move flush list retrieval to where it is used.

Commit Message

Sebastian Andrzej Siewior June 28, 2024, 10:18 a.m. UTC

The bpf_net_ctx_get_.*_flush_list() are used at the top of the function.
This means the variable is always assigned even if unused. By moving the
function to where it is used, it is possible to delay the initialisation
until it is unavoidable.
Not sure how much this gains in reality but by looking at bq_enqueue()
(in devmap.c) gcc pushes one register less to the stack. \o/.

 Move flush list retrieval to where it is used.

Signed-off-by: Sebastian Andrzej Siewior <[e-mail]>
---
 kernel/bpf/cpumap.c | 6 ++++--
 kernel/bpf/devmap.c | 3 ++-
 net/xdp/xsk.c       | 6 ++++--
 3 files changed, 10 insertions(+), 5 deletions(-)

Comments

Jesper Dangaard Brouer June 28, 2024, 4:39 p.m. UTC | #1

On 28/06/2024 12.18, Sebastian Andrzej Siewior wrote:
> The bpf_net_ctx_get_.*_flush_list() are used at the top of the function.
> This means the variable is always assigned even if unused. By moving the
> function to where it is used, it is possible to delay the initialisation
> until it is unavoidable.
> Not sure how much this gains in reality but by looking at bq_enqueue()
> (in devmap.c) gcc pushes one register less to the stack. \o/.
> 
>   Move flush list retrieval to where it is used.
> 
> Signed-off-by: Sebastian Andrzej Siewior<[e-mail]>
> ---
>   kernel/bpf/cpumap.c | 6 ++++--
>   kernel/bpf/devmap.c | 3 ++-
>   net/xdp/xsk.c       | 6 ++++--
>   3 files changed, 10 insertions(+), 5 deletions(-)

Nice small optimization :-)

Acked-by: Jesper Dangaard Brouer <[e-mail]>

Patch

diff --git a/kernel/bpf/cpumap.c b/kernel/bpf/cpumap.c
index 4acf90cd79eb4..fbdf5a1aabfe4 100644
--- a/kernel/bpf/cpumap.c
+++ b/kernel/bpf/cpumap.c
@@ -707,7 +707,6 @@  static void bq_flush_to_queue(struct xdp_bulk_queue *bq)
  */
 static void bq_enqueue(struct bpf_cpu_map_entry *rcpu, struct xdp_frame *xdpf)
 {
-	struct list_head *flush_list = bpf_net_ctx_get_cpu_map_flush_list();
 	struct xdp_bulk_queue *bq = this_cpu_ptr(rcpu->bulkq);
 
 	if (unlikely(bq->count == CPU_MAP_BULK_SIZE))
@@ -724,8 +723,11 @@  static void bq_enqueue(struct bpf_cpu_map_entry *rcpu, struct xdp_frame *xdpf)
 	 */
 	bq->q[bq->count++] = xdpf;
 
-	if (!bq->flush_node.prev)
+	if (!bq->flush_node.prev) {
+		struct list_head *flush_list = bpf_net_ctx_get_cpu_map_flush_list();
+
 		list_add(&bq->flush_node, flush_list);
+	}
 }
 
 int cpu_map_enqueue(struct bpf_cpu_map_entry *rcpu, struct xdp_frame *xdpf,
diff --git a/kernel/bpf/devmap.c b/kernel/bpf/devmap.c
index 9ca47eaacdd5e..b18d4a14a0a70 100644
--- a/kernel/bpf/devmap.c
+++ b/kernel/bpf/devmap.c
@@ -448,7 +448,6 @@  static void *__dev_map_lookup_elem(struct bpf_map *map, u32 key)
 static void bq_enqueue(struct net_device *dev, struct xdp_frame *xdpf,
 		       struct net_device *dev_rx, struct bpf_prog *xdp_prog)
 {
-	struct list_head *flush_list = bpf_net_ctx_get_dev_flush_list();
 	struct xdp_dev_bulk_queue *bq = this_cpu_ptr(dev->xdp_bulkq);
 
 	if (unlikely(bq->count == DEV_MAP_BULK_SIZE))
@@ -462,6 +461,8 @@  static void bq_enqueue(struct net_device *dev, struct xdp_frame *xdpf,
 	 * are only ever modified together.
 	 */
 	if (!bq->dev_rx) {
+		struct list_head *flush_list = bpf_net_ctx_get_dev_flush_list();
+
 		bq->dev_rx = dev_rx;
 		bq->xdp_prog = xdp_prog;
 		list_add(&bq->flush_node, flush_list);
diff --git a/net/xdp/xsk.c b/net/xdp/xsk.c
index de9c0322bc294..7e16336044b2d 100644
--- a/net/xdp/xsk.c
+++ b/net/xdp/xsk.c
@@ -370,15 +370,17 @@  static int xsk_rcv(struct xdp_sock *xs, struct xdp_buff *xdp)
 
 int __xsk_map_redirect(struct xdp_sock *xs, struct xdp_buff *xdp)
 {
-	struct list_head *flush_list = bpf_net_ctx_get_xskmap_flush_list();
 	int err;
 
 	err = xsk_rcv(xs, xdp);
 	if (err)
 		return err;
 
-	if (!xs->flush_node.prev)
+	if (!xs->flush_node.prev) {
+		struct list_head *flush_list = bpf_net_ctx_get_xskmap_flush_list();
+
 		list_add(&xs->flush_node, flush_list);
+	}
 
 	return 0;
 }