net/mlx5e: Fix inconsistent hairpin RQT sizes

The processing of traffic in hairpin queues occurs in HW/FW and does not
involve the cpus, hence the upper bound on max num channels does not
apply to them.  Using this bound for the hairpin RQT max_table_size is
wrong.  It could be too small, and cause the error below [1].  As the
RQT size provided on init does not get modified later, use the same
value for both actual and max table sizes.

[1]
mlx5_core 0000:08:00.1: mlx5_cmd_out_err:805:(pid 1200): CREATE_RQT(0x916) op_mod(0x0) failed, status bad parameter(0x3), syndrome (0x538faf), err(-22)

Fixes: 74a8dadac17e ("net/mlx5e: Preparations for supporting larger number of channels")
Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
Reviewed-by: Gal Pressman <gal@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>

diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en_tc.c b/drivers/net/ethernet/mellanox/mlx5/core/en_tc.c
index 30932c9c9a8f08bca2c8025f0a5685f79695e54d..047b465fc6a5348d39b868a53520475ddddcc0ac 100644 (file)
--- a/drivers/net/ethernet/mellanox/mlx5/core/en_tc.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en_tc.c
@@ -761,7 +761,7 @@ static int mlx5e_hairpin_create_indirect_rqt(struct mlx5e_hairpin *hp)
 
        err = mlx5e_rss_params_indir_init(&indir, mdev,
                                          mlx5e_rqt_size(mdev, hp->num_channels),
-                                         mlx5e_rqt_size(mdev, priv->max_nch));
+                                         mlx5e_rqt_size(mdev, hp->num_channels));
        if (err)
                return err;