[patch net-next] mlxsw: spectrum: Add FDB lock to prevent session interleaving

[patch net-next] mlxsw: spectrum: Add FDB lock to prevent session interleaving

[Jiri Pirko]

From: Ido Schimmel <idosch@mellanox.com>

Dumping the FDB (invoked with a process context) or handling FDB
notifications (polled periodicly in delayed work) might each entail
multiple EMAD transcations due to the number of entries.

While we only allow one EMAD transaction at a time, there is nothing
stopping the dump and notification processing sessions from
interleaving. However, this is forbidden by the hardware, so we need to
make sure only one of these sessions can run at a time.

Solve this by adding a mutex ('fdb_lock'), as both kernel threads can
sleep while waiting for the response EMAD.

Fixes: 56ade8fe3f ("mlxsw: spectrum: Add initial support for Spectrum ASIC")
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
---
 drivers/net/ethernet/mellanox/mlxsw/spectrum.h           | 1 +
 drivers/net/ethernet/mellanox/mlxsw/spectrum_switchdev.c | 5 +++++
 2 files changed, 6 insertions(+)

diff --git a/drivers/net/ethernet/mellanox/mlxsw/spectrum.h b/drivers/net/ethernet/mellanox/mlxsw/spectrum.h
index 7601789..0bbaef3 100644
--- a/drivers/net/ethernet/mellanox/mlxsw/spectrum.h
+++ b/drivers/net/ethernet/mellanox/mlxsw/spectrum.h
@@ -105,6 +105,7 @@ struct mlxsw_sp {
 	} fdb_notify;
 #define MLXSW_SP_DEFAULT_AGEING_TIME 300
 	u32 ageing_time;
+	struct mutex fdb_lock;	/* Make sure FDB sessions are atomic. */
 	struct mlxsw_sp_upper master_bridge;
 	struct mlxsw_sp_upper lags[MLXSW_SP_LAG_MAX];
 };
diff --git a/drivers/net/ethernet/mellanox/mlxsw/spectrum_switchdev.c b/drivers/net/ethernet/mellanox/mlxsw/spectrum_switchdev.c
index d64559e..94bc73c 100644
--- a/drivers/net/ethernet/mellanox/mlxsw/spectrum_switchdev.c
+++ b/drivers/net/ethernet/mellanox/mlxsw/spectrum_switchdev.c
@@ -881,6 +881,7 @@ static int mlxsw_sp_port_fdb_dump(struct mlxsw_sp_port *mlxsw_sp_port,
 		vport_vid = mlxsw_sp_vport_vid_get(mlxsw_sp_port);
 	}
 
+	mutex_lock(&mlxsw_sp->fdb_lock);
 	mlxsw_reg_sfd_pack(sfd_pl, MLXSW_REG_SFD_OP_QUERY_DUMP, 0);
 	do {
 		mlxsw_reg_sfd_num_rec_set(sfd_pl, MLXSW_REG_SFD_REC_MAX_COUNT);
@@ -938,6 +939,7 @@ static int mlxsw_sp_port_fdb_dump(struct mlxsw_sp_port *mlxsw_sp_port,
 	} while (num_rec == MLXSW_REG_SFD_REC_MAX_COUNT);
 
 out:
+	mutex_unlock(&mlxsw_sp->fdb_lock);
 	kfree(sfd_pl);
 	return stored_err ? stored_err : err;
 }
@@ -1187,6 +1189,7 @@ static void mlxsw_sp_fdb_notify_work(struct work_struct *work)
 
 	mlxsw_sp = container_of(work, struct mlxsw_sp, fdb_notify.dw.work);
 
+	mutex_lock(&mlxsw_sp->fdb_lock);
 	do {
 		mlxsw_reg_sfn_pack(sfn_pl);
 		err = mlxsw_reg_query(mlxsw_sp->core, MLXSW_REG(sfn), sfn_pl);
@@ -1199,6 +1202,7 @@ static void mlxsw_sp_fdb_notify_work(struct work_struct *work)
 			mlxsw_sp_fdb_notify_rec_process(mlxsw_sp, sfn_pl, i);
 
 	} while (num_rec);
+	mutex_unlock(&mlxsw_sp->fdb_lock);
 
 	kfree(sfn_pl);
 	mlxsw_sp_fdb_notify_work_schedule(mlxsw_sp);
@@ -1213,6 +1217,7 @@ static int mlxsw_sp_fdb_init(struct mlxsw_sp *mlxsw_sp)
 		dev_err(mlxsw_sp->bus_info->dev, "Failed to set default ageing time\n");
 		return err;
 	}
+	mutex_init(&mlxsw_sp->fdb_lock);
 	INIT_DELAYED_WORK(&mlxsw_sp->fdb_notify.dw, mlxsw_sp_fdb_notify_work);
 	mlxsw_sp->fdb_notify.interval = MLXSW_SP_DEFAULT_LEARNING_INTERVAL;
 	mlxsw_sp_fdb_notify_work_schedule(mlxsw_sp);
-- 
1.9.3