net/mlx5: Make RoCE and SR-IOV LAG modes explicit

With the introduction of SR-IOV LAG, checking whether LAG is active
is no longer good enough, since RoCE and SR-IOV LAG each entails
different behavior by both the core and infiniband drivers.

This patch introduces facilities to discern LAG type, in addition to
mlx5_lag_is_active(). These are implemented in such a way as to allow
more complex mode combinations in the future.

Signed-off-by: Aviv Heller <avivh@mellanox.com>
Reviewed-by: Roi Dayan <roid@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>

7 files changed, 77 insertions, 26 deletions

diff --git a/drivers/infiniband/hw/mlx5/main.c b/drivers/infiniband/hw/mlx5/main.c
index d66457e6ffba..e85974ab06c0 100644
--- a/drivers/infiniband/hw/mlx5/main.c
+++ b/drivers/infiniband/hw/mlx5/main.c
@@ -445,7 +445,7 @@ static int mlx5_query_port_roce(struct ib_device *device, u8 port_num,
 	if (!ndev)
 		goto out;
 
-	if (mlx5_lag_is_active(dev->mdev)) {
+	if (dev->lag_active) {
 		rcu_read_lock();
 		upper = netdev_master_upper_dev_get_rcu(ndev);
 		if (upper) {
@@ -1848,7 +1848,7 @@ static struct ib_ucontext *mlx5_ib_alloc_ucontext(struct ib_device *ibdev,
 	context->lib_caps = req.lib_caps;
 	print_lib_caps(dev, context->lib_caps);
 
-	if (mlx5_lag_is_active(dev->mdev)) {
+	if (dev->lag_active) {
 		u8 port = mlx5_core_native_port_num(dev->mdev);
 
 		atomic_set(&context->tx_port_affinity,
@@ -4841,7 +4841,7 @@ static int mlx5_eth_lag_init(struct mlx5_ib_dev *dev)
 	struct mlx5_flow_table *ft;
 	int err;
 
-	if (!ns || !mlx5_lag_is_active(mdev))
+	if (!ns || !mlx5_lag_is_roce(mdev))
 		return 0;
 
 	err = mlx5_cmd_create_vport_lag(mdev);
@@ -4855,6 +4855,7 @@ static int mlx5_eth_lag_init(struct mlx5_ib_dev *dev)
 	}
 
 	dev->flow_db->lag_demux_ft = ft;
+	dev->lag_active = true;
 	return 0;
 
 err_destroy_vport_lag:
@@ -4866,7 +4867,9 @@ static void mlx5_eth_lag_cleanup(struct mlx5_ib_dev *dev)
 {
 	struct mlx5_core_dev *mdev = dev->mdev;
 
-	if (dev->flow_db->lag_demux_ft) {
+	if (dev->lag_active) {
+		dev->lag_active = false;
+
 		mlx5_destroy_flow_table(dev->flow_db->lag_demux_ft);
 		dev->flow_db->lag_demux_ft = NULL;
 
@@ -6173,7 +6176,7 @@ int mlx5_ib_stage_ib_reg_init(struct mlx5_ib_dev *dev)
 	const char *name;
 
 	rdma_set_device_sysfs_group(&dev->ib_dev, &mlx5_attr_group);
-	if (!mlx5_lag_is_active(dev->mdev))
+	if (!mlx5_lag_is_roce(dev->mdev))
 		name = "mlx5_%d";
 	else
 		name = "mlx5_bond_%d";
diff --git a/drivers/infiniband/hw/mlx5/mlx5_ib.h b/drivers/infiniband/hw/mlx5/mlx5_ib.h
index c89b3b44b22e..e507b6eb7c09 100644
--- a/drivers/infiniband/hw/mlx5/mlx5_ib.h
+++ b/drivers/infiniband/hw/mlx5/mlx5_ib.h
@@ -936,6 +936,7 @@ struct mlx5_ib_dev {
 	struct mlx5_ib_delay_drop	delay_drop;
 	const struct mlx5_ib_profile	*profile;
 	struct mlx5_eswitch_rep		*rep;
+	int				lag_active;
 
 	struct mlx5_ib_lb_state		lb;
 	u8			umr_fence;
diff --git a/drivers/infiniband/hw/mlx5/qp.c b/drivers/infiniband/hw/mlx5/qp.c
index 3747cc681b18..a0e9ff763d42 100644
--- a/drivers/infiniband/hw/mlx5/qp.c
+++ b/drivers/infiniband/hw/mlx5/qp.c
@@ -3258,7 +3258,7 @@ static int __mlx5_ib_modify_qp(struct ib_qp *ibqp,
 		    (ibqp->qp_type == IB_QPT_RAW_PACKET) ||
 		    (ibqp->qp_type == IB_QPT_XRC_INI) ||
 		    (ibqp->qp_type == IB_QPT_XRC_TGT)) {
-			if (mlx5_lag_is_active(dev->mdev)) {
+			if (dev->lag_active) {
 				u8 p = mlx5_core_native_port_num(dev->mdev);
 				tx_affinity = get_tx_affinity(dev, pd, base, p);
 				context->flags |= cpu_to_be32(tx_affinity << 24);
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en/tc_tun.c b/drivers/net/ethernet/mellanox/mlx5/core/en/tc_tun.c
index d5d161ab0dbc..b92f8b3ff6b2 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en/tc_tun.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en/tc_tun.c
@@ -35,7 +35,7 @@ static int mlx5e_route_lookup_ipv4(struct mlx5e_priv *priv,
 	dst_is_lag_dev = (uplink_upper &&
 			  netif_is_lag_master(uplink_upper) &&
 			  rt->dst.dev == uplink_upper &&
-			  mlx5_lag_is_active(priv->mdev));
+			  mlx5_lag_is_sriov(priv->mdev));
 
 	/* if the egress device isn't on the same HW e-switch or
 	 * it's a LAG device, use the uplink
@@ -94,7 +94,7 @@ static int mlx5e_route_lookup_ipv6(struct mlx5e_priv *priv,
 	dst_is_lag_dev = (uplink_upper &&
 			  netif_is_lag_master(uplink_upper) &&
 			  dst->dev == uplink_upper &&
-			  mlx5_lag_is_active(priv->mdev));
+			  mlx5_lag_is_sriov(priv->mdev));
 
 	/* if the egress device isn't on the same HW e-switch or
 	 * it's a LAG device, use the uplink
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en_rep.c b/drivers/net/ethernet/mellanox/mlx5/core/en_rep.c
index 17f24127a3ba..e4a34c9ef700 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en_rep.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en_rep.c
@@ -314,7 +314,7 @@ int mlx5e_attr_get(struct net_device *dev, struct switchdev_attr *attr)
 	switch (attr->id) {
 	case SWITCHDEV_ATTR_ID_PORT_PARENT_ID:
 		attr->u.ppid.id_len = ETH_ALEN;
-		if (uplink_upper && mlx5_lag_is_active(uplink_priv->mdev)) {
+		if (uplink_upper && mlx5_lag_is_sriov(uplink_priv->mdev)) {
 			ether_addr_copy(attr->u.ppid.id, uplink_upper->dev_addr);
 		} else {
 			struct mlx5e_rep_priv *rpriv = priv->ppriv;
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en_tc.c b/drivers/net/ethernet/mellanox/mlx5/core/en_tc.c
index 864f3b00d09d..53ebb5a48018 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en_tc.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en_tc.c
@@ -2718,7 +2718,7 @@ static bool is_peer_flow_needed(struct mlx5e_tc_flow *flow)
 	bool esw_paired = mlx5_devcom_is_paired(attr->in_mdev->priv.devcom,
 						MLX5_DEVCOM_ESW_OFFLOADS);
 
-	return esw_paired && mlx5_lag_is_active(attr->in_mdev) &&
+	return esw_paired && mlx5_lag_is_sriov(attr->in_mdev) &&
 	       (is_rep_ingress || act_is_encap);
 }
 
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/lag.c b/drivers/net/ethernet/mellanox/mlx5/core/lag.c
index db5ef7023371..feb8230d3f86 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/lag.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/lag.c
@@ -37,9 +37,12 @@
 #include "eswitch.h"
 
 enum {
-	MLX5_LAG_FLAG_BONDED = 1 << 0,
+	MLX5_LAG_FLAG_ROCE   = 1 << 0,
+	MLX5_LAG_FLAG_SRIOV  = 1 << 1,
 };
 
+#define MLX5_LAG_MODE_FLAGS (MLX5_LAG_FLAG_ROCE | MLX5_LAG_FLAG_SRIOV)
+
 struct lag_func {
 	struct mlx5_core_dev *dev;
 	struct net_device    *netdev;
@@ -161,9 +164,19 @@ static int mlx5_lag_dev_get_netdev_idx(struct mlx5_lag *ldev,
 	return -1;
 }
 
+static bool __mlx5_lag_is_roce(struct mlx5_lag *ldev)
+{
+	return !!(ldev->flags & MLX5_LAG_FLAG_ROCE);
+}
+
+static bool __mlx5_lag_is_sriov(struct mlx5_lag *ldev)
+{
+	return !!(ldev->flags & MLX5_LAG_FLAG_SRIOV);
+}
+
 static bool __mlx5_lag_is_active(struct mlx5_lag *ldev)
 {
-	return !!(ldev->flags & MLX5_LAG_FLAG_BONDED);
+	return !!(ldev->flags & MLX5_LAG_MODE_FLAGS);
 }
 
 static void mlx5_infer_tx_affinity_mapping(struct lag_tracker *tracker,
@@ -229,9 +242,10 @@ static int mlx5_create_lag(struct mlx5_lag *ldev,
 }
 
 static void mlx5_activate_lag(struct mlx5_lag *ldev,
-			      struct lag_tracker *tracker)
+			      struct lag_tracker *tracker,
+			      u8 flags)
 {
-	ldev->flags |= MLX5_LAG_FLAG_BONDED;
+	ldev->flags |= flags;
 	mlx5_create_lag(ldev, tracker);
 }
 
@@ -240,7 +254,7 @@ static void mlx5_deactivate_lag(struct mlx5_lag *ldev)
 	struct mlx5_core_dev *dev0 = ldev->pf[0].dev;
 	int err;
 
-	ldev->flags &= ~MLX5_LAG_FLAG_BONDED;
+	ldev->flags &= ~MLX5_LAG_MODE_FLAGS;
 
 	err = mlx5_cmd_destroy_lag(dev0);
 	if (err)
@@ -263,15 +277,13 @@ static void mlx5_do_bond(struct mlx5_lag *ldev)
 {
 	struct mlx5_core_dev *dev0 = ldev->pf[0].dev;
 	struct mlx5_core_dev *dev1 = ldev->pf[1].dev;
-	bool do_bond, sriov_enabled;
 	struct lag_tracker tracker;
+	bool do_bond, roce_lag;
 	int i;
 
 	if (!dev0 || !dev1)
 		return;
 
-	sriov_enabled = mlx5_sriov_is_enabled(dev0) || mlx5_sriov_is_enabled(dev1);
-
 	mutex_lock(&lag_mutex);
 	tracker = ldev->tracker;
 	mutex_unlock(&lag_mutex);
@@ -279,28 +291,35 @@ static void mlx5_do_bond(struct mlx5_lag *ldev)
 	do_bond = tracker.is_bonded && mlx5_lag_check_prereq(ldev);
 
 	if (do_bond && !__mlx5_lag_is_active(ldev)) {
-		if (!sriov_enabled)
+		roce_lag = !mlx5_sriov_is_enabled(dev0) &&
+			   !mlx5_sriov_is_enabled(dev1);
+
+		if (roce_lag)
 			for (i = 0; i < MLX5_MAX_PORTS; i++)
 				mlx5_remove_dev_by_protocol(ldev->pf[i].dev,
 							    MLX5_INTERFACE_PROTOCOL_IB);
 
-		mlx5_activate_lag(ldev, &tracker);
+		mlx5_activate_lag(ldev, &tracker,
+				  roce_lag ? MLX5_LAG_FLAG_ROCE :
+				  MLX5_LAG_FLAG_SRIOV);
 
-		if (!sriov_enabled) {
+		if (roce_lag) {
 			mlx5_add_dev_by_protocol(dev0, MLX5_INTERFACE_PROTOCOL_IB);
 			mlx5_nic_vport_enable_roce(dev1);
 		}
 	} else if (do_bond && __mlx5_lag_is_active(ldev)) {
 		mlx5_modify_lag(ldev, &tracker);
 	} else if (!do_bond && __mlx5_lag_is_active(ldev)) {
-		if (!sriov_enabled) {
+		roce_lag = __mlx5_lag_is_roce(ldev);
+
+		if (roce_lag) {
 			mlx5_remove_dev_by_protocol(dev0, MLX5_INTERFACE_PROTOCOL_IB);
 			mlx5_nic_vport_disable_roce(dev1);
 		}
 
 		mlx5_deactivate_lag(ldev);
 
-		if (!sriov_enabled)
+		if (roce_lag)
 			for (i = 0; i < MLX5_MAX_PORTS; i++)
 				if (ldev->pf[i].dev)
 					mlx5_add_dev_by_protocol(ldev->pf[i].dev,
@@ -572,6 +591,20 @@ void mlx5_lag_remove(struct mlx5_core_dev *dev)
 	}
 }
 
+bool mlx5_lag_is_roce(struct mlx5_core_dev *dev)
+{
+	struct mlx5_lag *ldev;
+	bool res;
+
+	mutex_lock(&lag_mutex);
+	ldev = mlx5_lag_dev_get(dev);
+	res  = ldev && __mlx5_lag_is_roce(ldev);
+	mutex_unlock(&lag_mutex);
+
+	return res;
+}
+EXPORT_SYMBOL(mlx5_lag_is_roce);
+
 bool mlx5_lag_is_active(struct mlx5_core_dev *dev)
 {
 	struct mlx5_lag *ldev;
@@ -586,6 +619,20 @@ bool mlx5_lag_is_active(struct mlx5_core_dev *dev)
 }
 EXPORT_SYMBOL(mlx5_lag_is_active);
 
+bool mlx5_lag_is_sriov(struct mlx5_core_dev *dev)
+{
+	struct mlx5_lag *ldev;
+	bool res;
+
+	mutex_lock(&lag_mutex);
+	ldev = mlx5_lag_dev_get(dev);
+	res  = ldev && __mlx5_lag_is_sriov(ldev);
+	mutex_unlock(&lag_mutex);
+
+	return res;
+}
+EXPORT_SYMBOL(mlx5_lag_is_sriov);
+
 void mlx5_lag_update(struct mlx5_core_dev *dev)
 {
 	struct mlx5_lag *ldev;
@@ -609,7 +656,7 @@ struct net_device *mlx5_lag_get_roce_netdev(struct mlx5_core_dev *dev)
 	mutex_lock(&lag_mutex);
 	ldev = mlx5_lag_dev_get(dev);
 
-	if (!(ldev && __mlx5_lag_is_active(ldev)))
+	if (!(ldev && __mlx5_lag_is_roce(ldev)))
 		goto unlock;
 
 	if (ldev->tracker.tx_type == NETDEV_LAG_TX_TYPE_ACTIVEBACKUP) {
@@ -638,7 +685,7 @@ bool mlx5_lag_intf_add(struct mlx5_interface *intf, struct mlx5_priv *priv)
 		return true;
 
 	ldev = mlx5_lag_dev_get(dev);
-	if (!ldev || !__mlx5_lag_is_active(ldev) || ldev->pf[0].dev == dev)
+	if (!ldev || !__mlx5_lag_is_roce(ldev) || ldev->pf[0].dev == dev)
 		return true;
 
 	/* If bonded, we do not add an IB device for PF1. */
@@ -665,7 +712,7 @@ int mlx5_lag_query_cong_counters(struct mlx5_core_dev *dev,
 
 	mutex_lock(&lag_mutex);
 	ldev = mlx5_lag_dev_get(dev);
-	if (ldev && __mlx5_lag_is_active(ldev)) {
+	if (ldev && __mlx5_lag_is_roce(ldev)) {
 		num_ports = MLX5_MAX_PORTS;
 		mdev[0] = ldev->pf[0].dev;
 		mdev[1] = ldev->pf[1].dev;