From: Maxim Mikityanskiy <[email protected]>
mlx5e_select_queue compares num_tc_x_num_ch to real_num_tx_queues to
determine if HTB and/or PTP offloads are active. If they are, it
calculates netdev_pick_tx() % num_tc_x_num_ch to prevent it from
selecting HTB and PTP queues for regular traffic. However, before the
channels are first activated, num_tc_x_num_ch is zero. If
ndo_select_queue gets called at this point, the HTB/PTP check will pass,
and mlx5e_select_queue will attempt to take a modulo by num_tc_x_num_ch,
which equals to zero.
This commit fixes the bug by assigning num_tc_x_num_ch to a non-zero
value before registering the netdev.
Fixes: 214baf22870c ("net/mlx5e: Support HTB offload")
Reported-by: Jesper Dangaard Brouer <[email protected]>
Signed-off-by: Maxim Mikityanskiy <[email protected]>
Reviewed-by: Tariq Toukan <[email protected]>
Signed-off-by: Saeed Mahameed <[email protected]>
---
drivers/net/ethernet/mellanox/mlx5/core/en_main.c | 5 +++++
1 file changed, 5 insertions(+)
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en_main.c
b/drivers/net/ethernet/mellanox/mlx5/core/en_main.c
index c8b8249846a9..158f947a8503 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en_main.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en_main.c
@@ -4979,6 +4979,11 @@ void mlx5e_build_nic_params(struct mlx5e_priv *priv,
struct mlx5e_xsk *xsk, u16
priv->max_nch);
params->num_tc = 1;
+ /* Set an initial non-zero value, so that mlx5e_select_queue won't
+ * divide by zero if called before first activating channels.
+ */
+ priv->num_tc_x_num_ch = params->num_channels * params->num_tc;
+
/* SQ */
params->log_sq_size = is_kdump_kernel() ?
MLX5E_PARAMS_MINIMUM_LOG_SQ_SIZE :
--
2.30.2
