diff mbox series

[net-next,07/14] net/mlx5e: Optimize promiscuous mode

Message ID 20210126232419.175836-8-saeedm@nvidia.com (mailing list archive)
State Accepted
Delegated to: Netdev Maintainers
Headers show
Series [net-next,01/14] devlink: Add DMAC filter generic packet trap | expand

Checks

Context Check Description
netdev/cover_letter success Pull request
netdev/fixes_present success Link
netdev/patch_count success Link
netdev/tree_selection success Clearly marked for net-next
netdev/subject_prefix success Link
netdev/cc_maintainers warning 7 maintainers not CCed: roid@mellanox.com lariel@mellanox.com borisp@mellanox.com leon@kernel.org huyn@mellanox.com linux-rdma@vger.kernel.org paulb@mellanox.com
netdev/source_inline success Was 0 now: 0
netdev/verify_signedoff success Link
netdev/module_param success Was 0 now: 0
netdev/build_32bit success Errors and warnings before: 0 this patch: 0
netdev/kdoc success Errors and warnings before: 0 this patch: 0
netdev/verify_fixes success Link
netdev/checkpatch warning WARNING: line length of 87 exceeds 80 columns WARNING: line length of 88 exceeds 80 columns WARNING: line length of 91 exceeds 80 columns WARNING: line length of 92 exceeds 80 columns
netdev/build_allmodconfig_warn success Errors and warnings before: 0 this patch: 0
netdev/header_inline success Link
netdev/stable success Stable not CCed

Commit Message

Saeed Mahameed Jan. 26, 2021, 11:24 p.m. UTC
From: Aya Levin <ayal@nvidia.com>

Change steering flow to optimize traffic in promiscuous mode. On demand,
add a high priority table containing a catch-all rule. All incoming
packets are caught by this rule and steered directly to the TTC table.
Prior to this change, packets in promiscuous mode may suffer from up to
4 steering hops before reaching TTC table.
In addition, this patch will allow us adding a catch-all rule at the end
of MAC table to serve MAC trap, with no impact on promiscuous mode
performance.

Signed-off-by: Aya Levin <ayal@nvidia.com>
Reviewed-by: Moshe Shemesh <moshe@nvidia.com>
Reviewed-by: Maor Gottlieb <maorg@nvidia.com>
Reviewed-by: Tariq Toukan <tariqt@nvidia.com>
Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
---
 .../net/ethernet/mellanox/mlx5/core/en/fs.h   |  10 +-
 .../net/ethernet/mellanox/mlx5/core/en_fs.c   | 120 +++++++++++++-----
 .../net/ethernet/mellanox/mlx5/core/fs_core.c |   4 +-
 3 files changed, 100 insertions(+), 34 deletions(-)
diff mbox series

Patch

diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en/fs.h b/drivers/net/ethernet/mellanox/mlx5/core/en/fs.h
index 5749557749b0..abe57f032b2d 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en/fs.h
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en/fs.h
@@ -44,6 +44,11 @@  struct mlx5e_l2_rule {
 
 #define MLX5E_L2_ADDR_HASH_SIZE BIT(BITS_PER_BYTE)
 
+struct mlx5e_promisc_table {
+	struct mlx5e_flow_table	ft;
+	struct mlx5_flow_handle	*rule;
+};
+
 struct mlx5e_vlan_table {
 	struct mlx5e_flow_table		ft;
 	DECLARE_BITMAP(active_cvlans, VLAN_N_VID);
@@ -62,7 +67,6 @@  struct mlx5e_l2_table {
 	struct hlist_head          netdev_mc[MLX5E_L2_ADDR_HASH_SIZE];
 	struct mlx5e_l2_rule	   broadcast;
 	struct mlx5e_l2_rule	   allmulti;
-	struct mlx5e_l2_rule	   promisc;
 	bool                       broadcast_enabled;
 	bool                       allmulti_enabled;
 	bool                       promisc_enabled;
@@ -126,7 +130,8 @@  struct mlx5e_ttc_table {
 
 /* NIC prio FTS */
 enum {
-	MLX5E_VLAN_FT_LEVEL = 0,
+	MLX5E_PROMISC_FT_LEVEL,
+	MLX5E_VLAN_FT_LEVEL,
 	MLX5E_L2_FT_LEVEL,
 	MLX5E_TTC_FT_LEVEL,
 	MLX5E_INNER_TTC_FT_LEVEL,
@@ -241,6 +246,7 @@  struct mlx5e_flow_steering {
 	struct mlx5e_ethtool_steering   ethtool;
 #endif
 	struct mlx5e_tc_table           tc;
+	struct mlx5e_promisc_table      promisc;
 	struct mlx5e_vlan_table         vlan;
 	struct mlx5e_l2_table           l2;
 	struct mlx5e_ttc_table          ttc;
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en_fs.c b/drivers/net/ethernet/mellanox/mlx5/core/en_fs.c
index e02e5895703d..a2db550c982e 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en_fs.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en_fs.c
@@ -46,7 +46,6 @@  static void mlx5e_del_l2_flow_rule(struct mlx5e_priv *priv,
 enum {
 	MLX5E_FULLMATCH = 0,
 	MLX5E_ALLMULTI  = 1,
-	MLX5E_PROMISC   = 2,
 };
 
 enum {
@@ -596,6 +595,83 @@  static void mlx5e_handle_netdev_addr(struct mlx5e_priv *priv)
 	mlx5e_apply_netdev_addr(priv);
 }
 
+#define MLX5E_PROMISC_GROUP0_SIZE BIT(0)
+#define MLX5E_PROMISC_TABLE_SIZE MLX5E_PROMISC_GROUP0_SIZE
+
+static int mlx5e_add_promisc_rule(struct mlx5e_priv *priv)
+{
+	struct mlx5_flow_table *ft = priv->fs.promisc.ft.t;
+	struct mlx5_flow_destination dest = {};
+	struct mlx5_flow_handle **rule_p;
+	MLX5_DECLARE_FLOW_ACT(flow_act);
+	struct mlx5_flow_spec *spec;
+	int err = 0;
+
+	spec = kvzalloc(sizeof(*spec), GFP_KERNEL);
+	if (!spec)
+		return -ENOMEM;
+	dest.type = MLX5_FLOW_DESTINATION_TYPE_FLOW_TABLE;
+	dest.ft = priv->fs.ttc.ft.t;
+
+	rule_p = &priv->fs.promisc.rule;
+	*rule_p = mlx5_add_flow_rules(ft, spec, &flow_act, &dest, 1);
+	if (IS_ERR(*rule_p)) {
+		err = PTR_ERR(*rule_p);
+		*rule_p = NULL;
+		netdev_err(priv->netdev, "%s: add promiscuous rule failed\n", __func__);
+	}
+	kvfree(spec);
+	return err;
+}
+
+static int mlx5e_create_promisc_table(struct mlx5e_priv *priv)
+{
+	struct mlx5e_flow_table *ft = &priv->fs.promisc.ft;
+	struct mlx5_flow_table_attr ft_attr = {};
+	int err;
+
+	ft_attr.max_fte = MLX5E_PROMISC_TABLE_SIZE;
+	ft_attr.autogroup.max_num_groups = 1;
+	ft_attr.level = MLX5E_PROMISC_FT_LEVEL;
+	ft_attr.prio = MLX5E_NIC_PRIO;
+
+	ft->t = mlx5_create_auto_grouped_flow_table(priv->fs.ns, &ft_attr);
+	if (IS_ERR(ft->t)) {
+		err = PTR_ERR(ft->t);
+		netdev_err(priv->netdev, "fail to create promisc table err=%d\n", err);
+		return err;
+	}
+
+	err = mlx5e_add_promisc_rule(priv);
+	if (err)
+		goto err_destroy_promisc_table;
+
+	return 0;
+
+err_destroy_promisc_table:
+	mlx5_destroy_flow_table(ft->t);
+	ft->t = NULL;
+
+	return err;
+}
+
+static void mlx5e_del_promisc_rule(struct mlx5e_priv *priv)
+{
+	if (WARN(!priv->fs.promisc.rule, "Trying to remove non-existing promiscuous rule"))
+		return;
+	mlx5_del_flow_rules(priv->fs.promisc.rule);
+	priv->fs.promisc.rule = NULL;
+}
+
+static void mlx5e_destroy_promisc_table(struct mlx5e_priv *priv)
+{
+	if (WARN(!priv->fs.promisc.ft.t, "Trying to remove non-existing promiscuous table"))
+		return;
+	mlx5e_del_promisc_rule(priv);
+	mlx5_destroy_flow_table(priv->fs.promisc.ft.t);
+	priv->fs.promisc.ft.t = NULL;
+}
+
 void mlx5e_set_rx_mode_work(struct work_struct *work)
 {
 	struct mlx5e_priv *priv = container_of(work, struct mlx5e_priv,
@@ -615,14 +691,15 @@  void mlx5e_set_rx_mode_work(struct work_struct *work)
 	bool disable_allmulti  =  ea->allmulti_enabled  && !allmulti_enabled;
 	bool enable_broadcast  = !ea->broadcast_enabled &&  broadcast_enabled;
 	bool disable_broadcast =  ea->broadcast_enabled && !broadcast_enabled;
+	int err;
 
 	if (enable_promisc) {
-		if (!priv->channels.params.vlan_strip_disable)
+		err = mlx5e_create_promisc_table(priv);
+		if (err)
+			enable_promisc = false;
+		if (!priv->channels.params.vlan_strip_disable && !err)
 			netdev_warn_once(ndev,
 					 "S-tagged traffic will be dropped while C-tag vlan stripping is enabled\n");
-		mlx5e_add_l2_flow_rule(priv, &ea->promisc, MLX5E_PROMISC);
-		if (!priv->fs.vlan.cvlan_filter_disabled)
-			mlx5e_add_any_vid_rules(priv);
 	}
 	if (enable_allmulti)
 		mlx5e_add_l2_flow_rule(priv, &ea->allmulti, MLX5E_ALLMULTI);
@@ -635,11 +712,8 @@  void mlx5e_set_rx_mode_work(struct work_struct *work)
 		mlx5e_del_l2_flow_rule(priv, &ea->broadcast);
 	if (disable_allmulti)
 		mlx5e_del_l2_flow_rule(priv, &ea->allmulti);
-	if (disable_promisc) {
-		if (!priv->fs.vlan.cvlan_filter_disabled)
-			mlx5e_del_any_vid_rules(priv);
-		mlx5e_del_l2_flow_rule(priv, &ea->promisc);
-	}
+	if (disable_promisc)
+		mlx5e_destroy_promisc_table(priv);
 
 	ea->promisc_enabled   = promisc_enabled;
 	ea->allmulti_enabled  = allmulti_enabled;
@@ -1306,9 +1380,6 @@  static int mlx5e_add_l2_flow_rule(struct mlx5e_priv *priv,
 		mc_dmac[0] = 0x01;
 		mv_dmac[0] = 0x01;
 		break;
-
-	case MLX5E_PROMISC:
-		break;
 	}
 
 	ai->rule = mlx5_add_flow_rules(ft, spec, &flow_act, &dest, 1);
@@ -1324,13 +1395,11 @@  static int mlx5e_add_l2_flow_rule(struct mlx5e_priv *priv,
 	return err;
 }
 
-#define MLX5E_NUM_L2_GROUPS	   3
-#define MLX5E_L2_GROUP1_SIZE	   BIT(0)
-#define MLX5E_L2_GROUP2_SIZE	   BIT(15)
-#define MLX5E_L2_GROUP3_SIZE	   BIT(0)
+#define MLX5E_NUM_L2_GROUPS	   2
+#define MLX5E_L2_GROUP1_SIZE	   BIT(15)
+#define MLX5E_L2_GROUP2_SIZE	   BIT(0)
 #define MLX5E_L2_TABLE_SIZE	   (MLX5E_L2_GROUP1_SIZE +\
-				    MLX5E_L2_GROUP2_SIZE +\
-				    MLX5E_L2_GROUP3_SIZE)
+				    MLX5E_L2_GROUP2_SIZE)
 static int mlx5e_create_l2_table_groups(struct mlx5e_l2_table *l2_table)
 {
 	int inlen = MLX5_ST_SZ_BYTES(create_flow_group_in);
@@ -1353,20 +1422,11 @@  static int mlx5e_create_l2_table_groups(struct mlx5e_l2_table *l2_table)
 	mc = MLX5_ADDR_OF(create_flow_group_in, in, match_criteria);
 	mc_dmac = MLX5_ADDR_OF(fte_match_param, mc,
 			       outer_headers.dmac_47_16);
-	/* Flow Group for promiscuous */
-	MLX5_SET_CFG(in, start_flow_index, ix);
-	ix += MLX5E_L2_GROUP1_SIZE;
-	MLX5_SET_CFG(in, end_flow_index, ix - 1);
-	ft->g[ft->num_groups] = mlx5_create_flow_group(ft->t, in);
-	if (IS_ERR(ft->g[ft->num_groups]))
-		goto err_destroy_groups;
-	ft->num_groups++;
-
 	/* Flow Group for full match */
 	eth_broadcast_addr(mc_dmac);
 	MLX5_SET_CFG(in, match_criteria_enable, MLX5_MATCH_OUTER_HEADERS);
 	MLX5_SET_CFG(in, start_flow_index, ix);
-	ix += MLX5E_L2_GROUP2_SIZE;
+	ix += MLX5E_L2_GROUP1_SIZE;
 	MLX5_SET_CFG(in, end_flow_index, ix - 1);
 	ft->g[ft->num_groups] = mlx5_create_flow_group(ft->t, in);
 	if (IS_ERR(ft->g[ft->num_groups]))
@@ -1377,7 +1437,7 @@  static int mlx5e_create_l2_table_groups(struct mlx5e_l2_table *l2_table)
 	eth_zero_addr(mc_dmac);
 	mc_dmac[0] = 0x01;
 	MLX5_SET_CFG(in, start_flow_index, ix);
-	ix += MLX5E_L2_GROUP3_SIZE;
+	ix += MLX5E_L2_GROUP2_SIZE;
 	MLX5_SET_CFG(in, end_flow_index, ix - 1);
 	ft->g[ft->num_groups] = mlx5_create_flow_group(ft->t, in);
 	if (IS_ERR(ft->g[ft->num_groups]))
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/fs_core.c b/drivers/net/ethernet/mellanox/mlx5/core/fs_core.c
index b899539a0786..3dbd63b9845d 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/fs_core.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/fs_core.c
@@ -105,8 +105,8 @@ 
 #define ETHTOOL_PRIO_NUM_LEVELS 1
 #define ETHTOOL_NUM_PRIOS 11
 #define ETHTOOL_MIN_LEVEL (KERNEL_MIN_LEVEL + ETHTOOL_NUM_PRIOS)
-/* Vlan, mac, ttc, inner ttc, {aRFS/accel and esp/esp_err} */
-#define KERNEL_NIC_PRIO_NUM_LEVELS 6
+/* Promiscuous, Vlan, mac, ttc, inner ttc, {aRFS/accel and esp/esp_err} */
+#define KERNEL_NIC_PRIO_NUM_LEVELS 7
 #define KERNEL_NIC_NUM_PRIOS 1
 /* One more level for tc */
 #define KERNEL_MIN_LEVEL (KERNEL_NIC_PRIO_NUM_LEVELS + 1)