Skip to contentSkip to conversation

eth: mlx5: expose NETIF_F_NTUPLE when ARFS is compiled out

PATCH Applied Jakub Kicinski 3 participants 5 messages Jul 11, 2024 – Jul 13, 2024

Subsystems (4)
lore
Share & tools
lore

Jakub Kicinski

author maintainer
#0 lore raw

PATCH v2 1/1 eth: mlx5: expose NETIF_F_NTUPLE when ARFS is compiled out

ARFS depends on NTUPLE filters, but the inverse is not true. Drivers which don't support ARFS commonly still support NTUPLE filtering. mlx5 has a Kconfig option to disable ARFS (MLX5_EN_ARFS) and does not advertise NTUPLE filters as a feature at all when ARFS is compiled out. That's not correct, ntuple filters indeed still work just fine (as long as MLX5_EN_RXNFC is enabled).

This is needed to make the RSS test not skip all RSS context related testing.

---
v2:
- hard wire to on (not propagating Simon's review because of this)
v1: https://lore.kernel.org/[email protected]

---
 drivers/net/ethernet/mellanox/mlx5/core/en/fs.h     | 13 +++++++++++++
 .../net/ethernet/mellanox/mlx5/core/en_ethtool.c    |  2 +-
 drivers/net/ethernet/mellanox/mlx5/core/en_fs.c     |  5 ++---
 drivers/net/ethernet/mellanox/mlx5/core/en_main.c   |  8 +++++---
 .../net/ethernet/mellanox/mlx5/core/ipoib/ipoib.c   |  8 +++-----
 5 files changed, 24 insertions(+), 12 deletions(-)
drivers/net/ethernet/mellanox/mlx5/core/en/fs.h +13 -0
index 4d6225e0eec7..1e8b7d330701 100644
@@ -154,6 +154,19 @@ struct mlx5e_tc_table *mlx5e_fs_get_tc(struct mlx5e_flow_steering *fs);
 struct mlx5e_l2_table *mlx5e_fs_get_l2(struct mlx5e_flow_steering *fs);
 struct mlx5_flow_namespace *mlx5e_fs_get_ns(struct mlx5e_flow_steering *fs, bool egress);
 void mlx5e_fs_set_ns(struct mlx5e_flow_steering *fs, struct mlx5_flow_namespace *ns, bool egress);
+
+static inline bool mlx5e_fs_has_arfs(struct net_device *netdev)
+{
+	return IS_ENABLED(CONFIG_MLX5_EN_ARFS) &&
+		netdev->hw_features & NETIF_F_NTUPLE;
+}
+
+static inline bool mlx5e_fs_want_arfs(struct net_device *netdev)
+{
+	return IS_ENABLED(CONFIG_MLX5_EN_ARFS) &&
+		netdev->features & NETIF_F_NTUPLE;
+}
+
 #ifdef CONFIG_MLX5_EN_RXNFC
 struct mlx5e_ethtool_steering *mlx5e_fs_get_ethtool(struct mlx5e_flow_steering *fs);
 #endif
drivers/net/ethernet/mellanox/mlx5/core/en_ethtool.c +1 -1
index 3320f12ba2db..5582c93a62f1 100644
@@ -525,7 +525,7 @@ int mlx5e_ethtool_set_channels(struct mlx5e_priv *priv,
 
 	opened = test_bit(MLX5E_STATE_OPENED, &priv->state);
 
-	arfs_enabled = opened && (priv->netdev->features & NETIF_F_NTUPLE);
+	arfs_enabled = opened && mlx5e_fs_want_arfs(priv->netdev);
 	if (arfs_enabled)
 		mlx5e_arfs_disable(priv->fs);
 
drivers/net/ethernet/mellanox/mlx5/core/en_fs.c +2 -3
index 8c5b291a171f..05058710d2c7 100644
@@ -1307,8 +1307,7 @@ int mlx5e_create_flow_steering(struct mlx5e_flow_steering *fs,
 		return -EOPNOTSUPP;
 
 	mlx5e_fs_set_ns(fs, ns, false);
-	err = mlx5e_arfs_create_tables(fs, rx_res,
-				       !!(netdev->hw_features & NETIF_F_NTUPLE));
+	err = mlx5e_arfs_create_tables(fs, rx_res, mlx5e_fs_has_arfs(netdev));
 	if (err) {
 		fs_err(fs, "Failed to create arfs tables, err=%d\n", err);
 		netdev->hw_features &= ~NETIF_F_NTUPLE;
@@ -1355,7 +1354,7 @@ int mlx5e_create_flow_steering(struct mlx5e_flow_steering *fs,
 err_destroy_inner_ttc_table:
 	mlx5e_destroy_inner_ttc_table(fs);
 err_destroy_arfs_tables:
-	mlx5e_arfs_destroy_tables(fs, !!(netdev->hw_features & NETIF_F_NTUPLE));
+	mlx5e_arfs_destroy_tables(fs, mlx5e_fs_has_arfs(netdev));
 
 	return err;
 }
drivers/net/ethernet/mellanox/mlx5/core/en_main.c +5 -3
index ff335527c10a..6f686fabed44 100644
@@ -5556,8 +5556,10 @@ static void mlx5e_build_nic_netdev(struct net_device *netdev)
 #if IS_ENABLED(CONFIG_MLX5_CLS_ACT)
 		netdev->hw_features      |= NETIF_F_HW_TC;
 #endif
-#ifdef CONFIG_MLX5_EN_ARFS
+#if IS_ENABLED(CONFIG_MLX5_EN_ARFS)
 		netdev->hw_features	 |= NETIF_F_NTUPLE;
+#elif IS_ENABLED(CONFIG_MLX5_EN_RXNFC)
+		netdev->features	 |= NETIF_F_NTUPLE;
 #endif
 	}
 
@@ -5731,7 +5733,7 @@ static int mlx5e_init_nic_rx(struct mlx5e_priv *priv)
 err_tc_nic_cleanup:
 	mlx5e_tc_nic_cleanup(priv);
 err_destroy_flow_steering:
-	mlx5e_destroy_flow_steering(priv->fs, !!(priv->netdev->hw_features & NETIF_F_NTUPLE),
+	mlx5e_destroy_flow_steering(priv->fs, mlx5e_fs_has_arfs(priv->netdev),
 				    priv->profile);
 err_destroy_rx_res:
 	mlx5e_rx_res_destroy(priv->rx_res);
@@ -5747,7 +5749,7 @@ static void mlx5e_cleanup_nic_rx(struct mlx5e_priv *priv)
 {
 	mlx5e_accel_cleanup_rx(priv);
 	mlx5e_tc_nic_cleanup(priv);
-	mlx5e_destroy_flow_steering(priv->fs, !!(priv->netdev->hw_features & NETIF_F_NTUPLE),
+	mlx5e_destroy_flow_steering(priv->fs, mlx5e_fs_has_arfs(priv->netdev),
 				    priv->profile);
 	mlx5e_rx_res_destroy(priv->rx_res);
 	priv->rx_res = NULL;
drivers/net/ethernet/mellanox/mlx5/core/ipoib/ipoib.c +3 -5
index 8e0404c0d1ca..0979d672d47f 100644
@@ -372,7 +372,7 @@ static int mlx5i_create_flow_steering(struct mlx5e_priv *priv)
 
 	mlx5e_fs_set_ns(priv->fs, ns, false);
 	err = mlx5e_arfs_create_tables(priv->fs, priv->rx_res,
-				       !!(priv->netdev->hw_features & NETIF_F_NTUPLE));
+				       mlx5e_fs_has_arfs(priv->netdev));
 	if (err) {
 		netdev_err(priv->netdev, "Failed to create arfs tables, err=%d\n",
 			   err);
@@ -391,8 +391,7 @@ static int mlx5i_create_flow_steering(struct mlx5e_priv *priv)
 	return 0;
 
 err_destroy_arfs_tables:
-	mlx5e_arfs_destroy_tables(priv->fs,
-				  !!(priv->netdev->hw_features & NETIF_F_NTUPLE));
+	mlx5e_arfs_destroy_tables(priv->fs, mlx5e_fs_has_arfs(priv->netdev));
 
 	return err;
 }
@@ -400,8 +399,7 @@ static int mlx5i_create_flow_steering(struct mlx5e_priv *priv)
 static void mlx5i_destroy_flow_steering(struct mlx5e_priv *priv)
 {
 	mlx5e_destroy_ttc_table(priv->fs);
-	mlx5e_arfs_destroy_tables(priv->fs,
-				  !!(priv->netdev->hw_features & NETIF_F_NTUPLE));
+	mlx5e_arfs_destroy_tables(priv->fs, mlx5e_fs_has_arfs(priv->netdev));
 	mlx5e_ethtool_cleanup_steering(priv->fs);
 }
 
Signature · 1 line
2.45.2
Acked-by on v2 1/1
#1 lore raw
Quoted 96 lines · Jakub Kicinski

ARFS depends on NTUPLE filters, but the inverse is not true. Drivers which don't support ARFS commonly still support NTUPLE filtering. mlx5 has a Kconfig option to disable ARFS (MLX5_EN_ARFS) and does not advertise NTUPLE filters as a feature at all when ARFS is compiled out. That's not correct, ntuple filters indeed still work just fine (as long as MLX5_EN_RXNFC is enabled).

This is needed to make the RSS test not skip all RSS context related testing.

---
v2:
- hard wire to on (not propagating Simon's review because of this)
v1: https://lore.kernel.org/[email protected]

---

drivers/net/ethernet/mellanox/mlx5/core/en/fs.h     | 13 +++++++++++++
.../net/ethernet/mellanox/mlx5/core/en_ethtool.c    |  2 +-
drivers/net/ethernet/mellanox/mlx5/core/en_fs.c     |  5 ++---
drivers/net/ethernet/mellanox/mlx5/core/en_main.c   |  8 +++++---
.../net/ethernet/mellanox/mlx5/core/ipoib/ipoib.c   |  8 +++-----
5 files changed, 24 insertions(+), 12 deletions(-)
drivers/net/ethernet/mellanox/mlx5/core/en/fs.h +13 -0
index 4d6225e0eec7..1e8b7d330701 100644
@@ -154,6 +154,19 @@ struct mlx5e_tc_table *mlx5e_fs_get_tc(struct mlx5e_flow_steering *fs);
 struct mlx5e_l2_table *mlx5e_fs_get_l2(struct mlx5e_flow_steering *fs);
 struct mlx5_flow_namespace *mlx5e_fs_get_ns(struct mlx5e_flow_steering *fs, bool egress);
 void mlx5e_fs_set_ns(struct mlx5e_flow_steering *fs, struct mlx5_flow_namespace *ns, bool egress);
+
+static inline bool mlx5e_fs_has_arfs(struct net_device *netdev)
+{
+	return IS_ENABLED(CONFIG_MLX5_EN_ARFS) &&
+		netdev->hw_features & NETIF_F_NTUPLE;
+}
+
+static inline bool mlx5e_fs_want_arfs(struct net_device *netdev)
+{
+	return IS_ENABLED(CONFIG_MLX5_EN_ARFS) &&
+		netdev->features & NETIF_F_NTUPLE;
+}
+
 #ifdef CONFIG_MLX5_EN_RXNFC
 struct mlx5e_ethtool_steering *mlx5e_fs_get_ethtool(struct mlx5e_flow_steering *fs);
 #endif
drivers/net/ethernet/mellanox/mlx5/core/en_ethtool.c +1 -1
index 3320f12ba2db..5582c93a62f1 100644
@@ -525,7 +525,7 @@ int mlx5e_ethtool_set_channels(struct mlx5e_priv *priv,
 
 	opened = test_bit(MLX5E_STATE_OPENED, &priv->state);
 
-	arfs_enabled = opened && (priv->netdev->features & NETIF_F_NTUPLE);
+	arfs_enabled = opened && mlx5e_fs_want_arfs(priv->netdev);
 	if (arfs_enabled)
 		mlx5e_arfs_disable(priv->fs);
 
drivers/net/ethernet/mellanox/mlx5/core/en_fs.c +1 -2
index 8c5b291a171f..05058710d2c7 100644
@@ -1307,8 +1307,7 @@ int mlx5e_create_flow_steering(struct mlx5e_flow_steering *fs,
 		return -EOPNOTSUPP;
 
 	mlx5e_fs_set_ns(fs, ns, false);
-	err = mlx5e_arfs_create_tables(fs, rx_res,
-				       !!(netdev->hw_features & NETIF_F_NTUPLE));
+	err = mlx5e_arfs_create_tables(fs, rx_res, mlx5e_fs_has_arfs(netdev));
 	if (err) {
 		fs_err(fs, "Failed to create arfs tables, err=%d\n", err);
 		netdev->hw_features &= ~NETIF_F_NTUPLE;
@@ -1355,7 +1354,7 @@ int mlx5e_create_flow_steering(struct mlx5e_flow_steering *fs,

err_destroy_inner_ttc_table:

	mlx5e_destroy_inner_ttc_table(fs);

err_destroy_arfs_tables:

drivers/net/ethernet/mellanox/mlx5/core/en_fs.c +1 -1
-	mlx5e_arfs_destroy_tables(fs, !!(netdev->hw_features & NETIF_F_NTUPLE));
+	mlx5e_arfs_destroy_tables(fs, mlx5e_fs_has_arfs(netdev));
	return err;
}
drivers/net/ethernet/mellanox/mlx5/core/en_main.c +0 -0
index ff335527c10a..6f686fabed44 100644

drivers/net/ethernet/mellanox/mlx5/core/en_main.c static void mlx5e_build_nic_netdev(struct net_device *netdev) · line 5562

Quoted 8 lines
drivers/net/ethernet/mellanox/mlx5/core/en_main.c +3 -1
@@ -5556,8 +5556,10 @@ static void mlx5e_build_nic_netdev(struct net_device *netdev)
 #if IS_ENABLED(CONFIG_MLX5_CLS_ACT)
 		netdev->hw_features      |= NETIF_F_HW_TC;
 #endif
-#ifdef CONFIG_MLX5_EN_ARFS
+#if IS_ENABLED(CONFIG_MLX5_EN_ARFS)
 		netdev->hw_features	 |= NETIF_F_NTUPLE;
+#elif IS_ENABLED(CONFIG_MLX5_EN_RXNFC)
+		netdev->features	 |= NETIF_F_NTUPLE;

Why default ON when RXNFC and OFF when ARFS ?
Default should be off always, and this needs to be advertised in hw_features in both cases.

I think this should be

#if IS_ENABLED(ARFS) || IS_ENABLED(RXFNC)
	netdev->hw_features |= NTUPLE;

Otherwise LGTM

author maintainer
#2 lore raw
Quoted 9 lines · On Thu, 11 Jul 2024 16:05:29 -0700 Saeed Mahameed wrote:

drivers/net/ethernet/mellanox/mlx5/core/en_main.c static void mlx5e_build_nic_netdev(struct net_device *netdev) · line 5562

Quoted 4 lines

+#if IS_ENABLED(CONFIG_MLX5_EN_ARFS)

		netdev->hw_features	 |= NETIF_F_NTUPLE;
drivers/net/ethernet/mellanox/mlx5/core/en_main.c +2 -0
+#elif IS_ENABLED(CONFIG_MLX5_EN_RXNFC)
+		netdev->features	 |= NETIF_F_NTUPLE;  

Why default ON when RXNFC and OFF when ARFS ?
Default should be off always, and this needs to be advertised in hw_features in both cases.

Quoted 3 lines

I think this should be

#if IS_ENABLED(ARFS) || IS_ENABLED(RXFNC)
	netdev->hw_features |= NTUPLE;

That's what I thought, but on reflection since the filters can be added, and disable doesn't actually do anything - "fixed on" started to sound more appropriate. The additional "[fixed]" could also be useful for troubleshooting, to signal that this is a different situation than ARFS=y. No hard preference tho.

Quoted 3 lines

Otherwise LGTM

Thanks!

#3 lore raw
Quoted 14 lines · Jakub Kicinski
Quoted 9 lines · On Thu, 11 Jul 2024 16:05:29 -0700 Saeed Mahameed wrote:

drivers/net/ethernet/mellanox/mlx5/core/en_main.c static void mlx5e_build_nic_netdev(struct net_device *netdev) · line 5562

Quoted 4 lines

+#if IS_ENABLED(CONFIG_MLX5_EN_ARFS)

		netdev->hw_features	 |= NETIF_F_NTUPLE;
drivers/net/ethernet/mellanox/mlx5/core/en_main.c +2 -0
+#elif IS_ENABLED(CONFIG_MLX5_EN_RXNFC)
+		netdev->features	 |= NETIF_F_NTUPLE;

Why default ON when RXNFC and OFF when ARFS ?
Default should be off always, and this needs to be advertised in hw_features in both cases.

Quoted 3 lines

I think this should be

#if IS_ENABLED(ARFS) || IS_ENABLED(RXFNC)
	netdev->hw_features |= NTUPLE;
Quoted 6 lines

That's what I thought, but on reflection since the filters can be added, and disable doesn't actually do anything - "fixed on" started to sound more appropriate. The additional "[fixed]" could also be useful for troubleshooting, to signal that this is a different situation than ARFS=y. No hard preference tho.

Agreed, Thanks.

Quoted 5 lines
Quoted 3 lines

Otherwise LGTM

Thanks!

#4 lore raw
In reply to Jakub Kicinski
This patch was applied to netdev/net-next.git (main) by Jakub Kicinski <[email protected]>:

Hello:

This patch was applied to netdev/net-next.git (main) by Jakub Kicinski <[email protected]>:

Quoted 8 lines · On Thu, 11 Jul 2024 15:37:22 -0700 you wrote:

ARFS depends on NTUPLE filters, but the inverse is not true. Drivers which don't support ARFS commonly still support NTUPLE filtering. mlx5 has a Kconfig option to disable ARFS (MLX5_EN_ARFS) and does not advertise NTUPLE filters as a feature at all when ARFS is compiled out. That's not correct, ntuple filters indeed still work just fine (as long as MLX5_EN_RXNFC is enabled).

[...]

Here is the summary with links:
- [net-next,v2] eth: mlx5: expose NETIF_F_NTUPLE when ARFS is compiled out

    https://git.kernel.org/netdev/net-next/c/3771266bf841

You are awesome, thank you!

Signature · 2 lines
Deet-doot-dot, I am a bot.
https://korg.docs.kernel.org/patchwork/pwbot.html