net/mlx5e: resize non-default RSS indirection tables on channel change

When the channel count changes and the RQT size changes with it, a
problem arise for non-default RSS contexts. The driver-side indirection
table grows actual_table_size without filling the new entries; stale
entries from a prior larger configuration may be re-exposed, causing
mlx5e_calc_indir_rqns() to WARN on an out-of-range index.

Replace mlx5e_rss_params_indir_modify_actual_size() with
mlx5e_rss_ctx_resize(), which fills new entries by replicating
the existing pattern, matching what ethtool_rxfh_ctxs_resize() does
for the same case. And restrict the loop to non-default contexts.

Call ethtool_rxfh_ctxs_can_resize() before acquiring state_lock to
validate that all non-default contexts can be resized, and
ethtool_rxfh_ctxs_resize() after releasing it to fold or unfold their
indirection tables. Both functions acquire rss_lock internally and
cannot be called under state_lock. RTNL, held by all set_channels
callers, serialises context creation and deletion making the pre-lock
check safe.

Guard both ethtool calls on mlx5e_rx_res_rss_cnt() > 1: skip the
validation and resize when no non-default contexts exist. This
naturally covers representors and IPoIB, which share
mlx5e_ethtool_set_channels() but cannot have non-default RSS contexts.

Signed-off-by: Yael Chemla <ychemla@nvidia.com>
Reviewed-by: Nimrod Oren <noren@nvidia.com>
Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
Link: https://patch.msgid.link/20260511172719.330490-4-tariqt@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

drivers/net/ethernet/mellanox/mlx5/core/en/rss.c

@@ -85,9 +85,21 @@ bool mlx5e_rss_get_inner_ft_support(struct mlx5e_rss *rss)
 	return rss->params.inner_ft_support;
 }
 
-void mlx5e_rss_params_indir_modify_actual_size(struct mlx5e_rss *rss, u32 num_channels)
+void mlx5e_rss_set_indir_actual_size(struct mlx5e_rss *rss, u32 size)
 {
-	rss->indir.actual_table_size = mlx5e_rqt_size(rss->mdev, num_channels);
+	rss->indir.actual_table_size = size;
+}
+
+/* Handles non-default contexts, replicate existing pattern into new entries,
+ * matching what ethtool_rxfh_ctxs_resize() does.
+ */
+void mlx5e_rss_ctx_resize(struct mlx5e_rss *rss, u32 new_size)
+{
+	u32 old_size = rss->indir.actual_table_size;
+	u32 i;
+
+	for (i = old_size; i < new_size; i++)
+		rss->indir.table[i] = rss->indir.table[i % old_size];
 }
 
 int mlx5e_rss_params_indir_init(struct mlx5e_rss_params_indir *indir,

drivers/net/ethernet/mellanox/mlx5/core/en/rss.h

@@ -34,7 +34,7 @@ struct mlx5e_rss;
 int mlx5e_rss_params_indir_init(struct mlx5e_rss_params_indir *indir,
 				u32 actual_table_size, u32 max_table_size);
 void mlx5e_rss_params_indir_cleanup(struct mlx5e_rss_params_indir *indir);
-void mlx5e_rss_params_indir_modify_actual_size(struct mlx5e_rss *rss, u32 num_channels);
+void mlx5e_rss_ctx_resize(struct mlx5e_rss *rss, u32 new_size);
 struct mlx5e_rss *
 mlx5e_rss_init(struct mlx5_core_dev *mdev,
 	       const struct mlx5e_rss_params *params,
@@ -46,6 +46,7 @@ void mlx5e_rss_refcnt_dec(struct mlx5e_rss *rss);
 unsigned int mlx5e_rss_refcnt_read(struct mlx5e_rss *rss);
 
 bool mlx5e_rss_get_inner_ft_support(struct mlx5e_rss *rss);
+void mlx5e_rss_set_indir_actual_size(struct mlx5e_rss *rss, u32 size);
 u32 mlx5e_rss_get_tirn(struct mlx5e_rss *rss, enum mlx5_traffic_types tt,
 		       bool inner);
 bool mlx5e_rss_valid_tir(struct mlx5e_rss *rss, enum mlx5_traffic_types tt, bool inner);

drivers/net/ethernet/mellanox/mlx5/core/en/rx_res.c

@@ -42,11 +42,20 @@ static u32 *get_vhca_ids(struct mlx5e_rx_res *res, int offset)
 
 void mlx5e_rx_res_rss_update_num_channels(struct mlx5e_rx_res *res, u32 nch)
 {
+	u32 new_size = mlx5e_rqt_size(res->mdev, nch);
 	int i;
 
-	for (i = 0; i < MLX5E_MAX_NUM_RSS; i++) {
-		if (res->rss[i])
-			mlx5e_rss_params_indir_modify_actual_size(res->rss[i], nch);
+	WARN_ON_ONCE(res->rss_active);
+
+	/* Default context */
+	mlx5e_rss_set_indir_actual_size(res->rss[0], new_size);
+
+	/* Non-default contexts */
+	for (i = 1; i < MLX5E_MAX_NUM_RSS; i++) {
+		if (res->rss[i]) {
+			mlx5e_rss_ctx_resize(res->rss[i], new_size);
+			mlx5e_rss_set_indir_actual_size(res->rss[i], new_size);
+		}
 	}
 }
 

drivers/net/ethernet/mellanox/mlx5/core/en_ethtool.c

@@ -499,11 +499,15 @@ int mlx5e_ethtool_set_channels(struct mlx5e_priv *priv,
 {
 	struct mlx5e_params *cur_params = &priv->channels.params;
 	unsigned int count = ch->combined_count;
+	int new_rqt_size, cur_rqt_size;
 	struct mlx5e_params new_params;
 	bool arfs_enabled;
+	bool has_rss_ctxs;
 	bool opened;
 	int err = 0;
 
+	ASSERT_RTNL();
+
 	if (!count) {
 		netdev_info(priv->netdev, "%s: combined_count=0 not supported\n",
 			    __func__);
@@ -513,16 +517,33 @@ int mlx5e_ethtool_set_channels(struct mlx5e_priv *priv,
 	if (cur_params->num_channels == count)
 		return 0;
 
+	new_rqt_size = mlx5e_rqt_size(priv->mdev, count);
+	/* Validate that all non-default RSS contexts can be resized before
+	 * committing to the channel count change.
+	 * ethtool_rxfh_ctxs_can_resize() acquires rss_lock internally and
+	 * cannot be called under state_lock (rss_lock -> state_lock ordering).
+	 */
+	has_rss_ctxs = priv->rx_res && mlx5e_rx_res_rss_cnt(priv->rx_res) > 1;
+	if (has_rss_ctxs) {
+		err = ethtool_rxfh_ctxs_can_resize(priv->netdev, new_rqt_size);
+		if (err)
+			return err;
+	}
+
 	mutex_lock(&priv->state_lock);
 
+	if (!priv->rx_res) {
+		err = -EINVAL;
+		goto out;
+	}
+
+	cur_rqt_size = mlx5e_rqt_size(priv->mdev, cur_params->num_channels);
+
 	/* If RXFH is configured, changing the channels number is allowed only if
 	 * it does not require resizing the RSS table. This is because the previous
 	 * configuration may no longer be compatible with the new RSS table.
 	 */
 	if (netif_is_rxfh_configured(priv->netdev)) {
-		int cur_rqt_size = mlx5e_rqt_size(priv->mdev, cur_params->num_channels);
-		int new_rqt_size = mlx5e_rqt_size(priv->mdev, count);
-
 		if (new_rqt_size != cur_rqt_size) {
 			err = -EINVAL;
 			netdev_err(priv->netdev,
@@ -577,6 +598,14 @@ int mlx5e_ethtool_set_channels(struct mlx5e_priv *priv,
 out:
 	mutex_unlock(&priv->state_lock);
 
+	/* After a successful channel count change that altered the RQT size,
+	 * fold or unfold the indirection tables of all non-default RSS
+	 * contexts. Must run after state_lock is released because
+	 * ethtool_rxfh_ctxs_resize() acquires rss_lock internally.
+	 */
+	if (!err && cur_rqt_size != new_rqt_size && has_rss_ctxs)
+		ethtool_rxfh_ctxs_resize(priv->netdev, new_rqt_size);
+
 	return err;
 }