From: Moni Shoua <monis@mellanox.com>
Date: Tue, 9 May 2017 12:20:55 +0300
Subject: net/mlx5: Undo LAG upon request to create virtual functions
Patch-mainline: v4.13-rc1
Git-commit: 552db7bca568a8efbd0883f0597fbc42333cf094
References: bsc#1046303 FATE#322944
LAG cannot work if virtual functions are present. Therefore, if LAG is
configured, the attempt to create virtual functions will fail. This gives
precedence to LAG over SRIOV which is not the desired behavior as users
might want to use the bonding/teaming driver also want to work with SRIOV.
In that case we don't want to force an order of actions, first create
virtual functions and only than configure a bonding/teaming net device.
To fix, if LAG is configured during a request to create virtual
functions, remove it and continue.
We ignore ENODEV when trying to forbid lag. This makes sense
because "No such device" means that lag is forbidden anyway.
Signed-off-by: Moni Shoua <monis@mellanox.com>
Reviewed-by: Aviv Heller <avivh@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Acked-by: Thomas Bogendoerfer <tbogendoerfer@suse.de>
---
drivers/net/ethernet/mellanox/mlx5/core/lag.c | 69 +++++++++++++++++---
drivers/net/ethernet/mellanox/mlx5/core/mlx5_core.h | 3
drivers/net/ethernet/mellanox/mlx5/core/sriov.c | 15 ++--
3 files changed, 74 insertions(+), 13 deletions(-)
--- a/drivers/net/ethernet/mellanox/mlx5/core/lag.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/lag.c
@@ -61,6 +61,11 @@ struct mlx5_lag {
struct lag_tracker tracker;
struct delayed_work bond_work;
struct notifier_block nb;
+
+ /* Admin state. Allow lag only if allowed is true
+ * even if network conditions for lag were met
+ */
+ bool allowed;
};
/* General purpose, use for short periods of time.
@@ -209,6 +214,7 @@ static void mlx5_do_bond(struct mlx5_lag
struct lag_tracker tracker;
u8 v2p_port1, v2p_port2;
int i, err;
+ bool do_bond;
if (!dev0 || !dev1)
return;
@@ -217,13 +223,9 @@ static void mlx5_do_bond(struct mlx5_lag
tracker = ldev->tracker;
mutex_unlock(&lag_mutex);
- if (tracker.is_bonded && !mlx5_lag_is_bonded(ldev)) {
- if (mlx5_sriov_is_enabled(dev0) ||
- mlx5_sriov_is_enabled(dev1)) {
- mlx5_core_warn(dev0, "LAG is not supported with SRIOV");
- return;
- }
+ do_bond = tracker.is_bonded && ldev->allowed;
+ if (do_bond && !mlx5_lag_is_bonded(ldev)) {
for (i = 0; i < MLX5_MAX_PORTS; i++)
mlx5_remove_dev_by_protocol(ldev->pf[i].dev,
MLX5_INTERFACE_PROTOCOL_IB);
@@ -232,7 +234,7 @@ static void mlx5_do_bond(struct mlx5_lag
mlx5_add_dev_by_protocol(dev0, MLX5_INTERFACE_PROTOCOL_IB);
mlx5_nic_vport_enable_roce(dev1);
- } else if (tracker.is_bonded && mlx5_lag_is_bonded(ldev)) {
+ } else if (do_bond && mlx5_lag_is_bonded(ldev)) {
mlx5_infer_tx_affinity_mapping(&tracker, &v2p_port1,
&v2p_port2);
@@ -247,7 +249,7 @@ static void mlx5_do_bond(struct mlx5_lag
"Failed to modify LAG (%d)\n",
err);
}
- } else if (!tracker.is_bonded && mlx5_lag_is_bonded(ldev)) {
+ } else if (!do_bond && mlx5_lag_is_bonded(ldev)) {
mlx5_remove_dev_by_protocol(dev0, MLX5_INTERFACE_PROTOCOL_IB);
mlx5_nic_vport_disable_roce(dev1);
@@ -406,6 +408,15 @@ static int mlx5_lag_netdev_event(struct
return NOTIFY_DONE;
}
+static bool mlx5_lag_check_prereq(struct mlx5_lag *ldev)
+{
+ if ((ldev->pf[0].dev && mlx5_sriov_is_enabled(ldev->pf[0].dev)) ||
+ (ldev->pf[1].dev && mlx5_sriov_is_enabled(ldev->pf[1].dev)))
+ return false;
+ else
+ return true;
+}
+
static struct mlx5_lag *mlx5_lag_dev_alloc(void)
{
struct mlx5_lag *ldev;
@@ -415,6 +426,7 @@ static struct mlx5_lag *mlx5_lag_dev_all
return NULL;
INIT_DELAYED_WORK(&ldev->bond_work, mlx5_do_bond_work);
+ ldev->allowed = mlx5_lag_check_prereq(ldev);
return ldev;
}
@@ -439,7 +451,9 @@ static void mlx5_lag_dev_add_pf(struct m
ldev->tracker.netdev_state[fn].link_up = 0;
ldev->tracker.netdev_state[fn].tx_enabled = 0;
+ ldev->allowed = mlx5_lag_check_prereq(ldev);
dev->priv.lag = ldev;
+
mutex_unlock(&lag_mutex);
}
@@ -459,6 +473,7 @@ static void mlx5_lag_dev_remove_pf(struc
memset(&ldev->pf[i], 0, sizeof(*ldev->pf));
dev->priv.lag = NULL;
+ ldev->allowed = mlx5_lag_check_prereq(ldev);
mutex_unlock(&lag_mutex);
}
@@ -537,6 +552,44 @@ bool mlx5_lag_is_active(struct mlx5_core
}
EXPORT_SYMBOL(mlx5_lag_is_active);
+static int mlx5_lag_set_state(struct mlx5_core_dev *dev, bool allow)
+{
+ struct mlx5_lag *ldev;
+ int ret = 0;
+ bool lag_active;
+
+ mlx5_dev_list_lock();
+
+ ldev = mlx5_lag_dev_get(dev);
+ if (!ldev) {
+ ret = -ENODEV;
+ goto unlock;
+ }
+ lag_active = mlx5_lag_is_bonded(ldev);
+ if (!mlx5_lag_check_prereq(ldev) && allow) {
+ ret = -EINVAL;
+ goto unlock;
+ }
+ if (ldev->allowed == allow)
+ goto unlock;
+ ldev->allowed = allow;
+ if ((lag_active && !allow) || allow)
+ mlx5_do_bond(ldev);
+unlock:
+ mlx5_dev_list_unlock();
+ return ret;
+}
+
+int mlx5_lag_forbid(struct mlx5_core_dev *dev)
+{
+ return mlx5_lag_set_state(dev, false);
+}
+
+int mlx5_lag_allow(struct mlx5_core_dev *dev)
+{
+ return mlx5_lag_set_state(dev, true);
+}
+
struct net_device *mlx5_lag_get_roce_netdev(struct mlx5_core_dev *dev)
{
struct net_device *ndev = NULL;
--- a/drivers/net/ethernet/mellanox/mlx5/core/mlx5_core.h
+++ b/drivers/net/ethernet/mellanox/mlx5/core/mlx5_core.h
@@ -172,4 +172,7 @@ static inline int mlx5_lag_is_lacp_owner
MLX5_CAP_GEN(dev, lag_master);
}
+int mlx5_lag_allow(struct mlx5_core_dev *dev);
+int mlx5_lag_forbid(struct mlx5_core_dev *dev);
+
#endif /* __MLX5_CORE_H__ */
--- a/drivers/net/ethernet/mellanox/mlx5/core/sriov.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/sriov.c
@@ -175,15 +175,20 @@ int mlx5_core_sriov_configure(struct pci
if (!mlx5_core_is_pf(dev))
return -EPERM;
- if (num_vfs && mlx5_lag_is_active(dev)) {
- mlx5_core_warn(dev, "can't turn sriov on while LAG is active");
- return -EINVAL;
+ if (num_vfs) {
+ int ret;
+
+ ret = mlx5_lag_forbid(dev);
+ if (ret && (ret != -ENODEV))
+ return ret;
}
- if (num_vfs)
+ if (num_vfs) {
err = mlx5_sriov_enable(pdev, num_vfs);
- else
+ } else {
mlx5_sriov_disable(pdev);
+ mlx5_lag_allow(dev);
+ }
return err ? err : num_vfs;
}