[SRU][N:linux-bluefield][PATCH v1 03/10] net/mlx5e: set the tx_queue_len for pfifo_fast

Stav Aviram saviram at nvidia.com
Mon Oct 6 09:31:49 UTC 2025


From: William Tu <witu at nvidia.com>

BugLink: https://bugs.launchpad.net/bugs/2126876

By default, the mq netdev creates a pfifo_fast qdisc. On a
system with 16 core, the pfifo_fast with 3 bands consumes
16 * 3 * 8 (size of pointer) * 1024 (default tx queue len)
= 393KB. The patch sets the tx qlen to representor default
value, 128 (1<<MLX5E_REP_PARAMS_DEF_LOG_SQ_SIZE), which
consumes 16 * 3 * 8 * 128 = 49KB, saving 344KB for each
representor at ECPF.

Signed-off-by: William Tu <witu at nvidia.com>
Reviewed-by: Daniel Jurgens <danielj at nvidia.com>
Signed-off-by: Tariq Toukan <tariqt at nvidia.com>
Reviewed-by: Michal Swiatkowski <michal.swiatkowski at linux.intel.com>
Link: https://patch.msgid.link/20250209101716.112774-9-tariqt@nvidia.com
Signed-off-by: Jakub Kicinski <kuba at kernel.org>
(cherry picked from commit a38cc5706fb9f7dc4ee3a443f61de13ce1e410ed)
Signed-off-by: Stav Aviram <saviram at nvidia.com>
---
 drivers/net/ethernet/mellanox/mlx5/core/en_rep.c | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en_rep.c b/drivers/net/ethernet/mellanox/mlx5/core/en_rep.c
index dc0836d58519..edf98dfd3628 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en_rep.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en_rep.c
@@ -856,6 +856,8 @@ static void mlx5e_build_rep_netdev(struct net_device *netdev,
 	netdev->ethtool_ops = &mlx5e_rep_ethtool_ops;
 
 	netdev->watchdog_timeo    = 15 * HZ;
+	if (mlx5_core_is_ecpf(mdev))
+		netdev->tx_queue_len = 1 << MLX5E_REP_PARAMS_DEF_LOG_SQ_SIZE;
 
 #if IS_ENABLED(CONFIG_MLX5_CLS_ACT)
 	netdev->hw_features    |= NETIF_F_HW_TC;
-- 
2.38.1




More information about the kernel-team mailing list