[PATCH 23.11 3/3] net/mlx5: fix rollback on failed flow configure

[Dariusz Sosnowski]

[ upstream commit 727283742a372359241fafec07f3648d3261e763 ]

If rte_flow_configure() failed, then some port resources
were either not freed, nor reset to the default state.
As a result, assumptions in other places in PMD were invalidated
and that lead to segmentation faults during release of HW Steering
resources when port was closed.

This patch adds missing resource release to rollback procedure
in mlx5 PMD implementation of rte_flow_configure().
Whole rollback procedure is reordered for clarity, to resemble
reverse order of resource allocation.

Fixes: 1939eb6f660c ("net/mlx5: support flow port action with HWS")
Fixes: 773ca0e91ba1 ("net/mlx5: support VLAN push/pop/modify with HWS")
Fixes: 04a4de756e14 ("net/mlx5: support flow age action with HWS")
Cc: stable at dpdk.org

Signed-off-by: Dariusz Sosnowski <dsosnowski at nvidia.com>
Acked-by: Ori Kam <orika at nvidia.com>
---
 drivers/net/mlx5/mlx5_flow_hw.c | 53 +++++++++++++++++++++------------
 1 file changed, 34 insertions(+), 19 deletions(-)

diff --git a/drivers/net/mlx5/mlx5_flow_hw.c b/drivers/net/mlx5/mlx5_flow_hw.c
index b51a4bb837..2f4b2d774e 100644
--- a/drivers/net/mlx5/mlx5_flow_hw.c
+++ b/drivers/net/mlx5/mlx5_flow_hw.c
@@ -9705,6 +9705,14 @@ flow_hw_configure(struct rte_eth_dev *dev,
 		priv->hws_strict_queue = 1;
 	return 0;
 err:
+	priv->hws_strict_queue = 0;
+	flow_hw_destroy_vlan(dev);
+	if (priv->hws_age_req)
+		mlx5_hws_age_pool_destroy(priv);
+	if (priv->hws_cpool) {
+		mlx5_hws_cnt_pool_destroy(priv->sh, priv->hws_cpool);
+		priv->hws_cpool = NULL;
+	}
 	if (priv->hws_ctpool) {
 		flow_hw_ct_pool_destroy(dev, priv->hws_ctpool);
 		priv->hws_ctpool = NULL;
@@ -9713,27 +9721,37 @@ flow_hw_configure(struct rte_eth_dev *dev,
 		flow_hw_ct_mng_destroy(dev, priv->ct_mng);
 		priv->ct_mng = NULL;
 	}
-	if (priv->hws_age_req)
-		mlx5_hws_age_pool_destroy(priv);
-	if (priv->hws_cpool) {
-		mlx5_hws_cnt_pool_destroy(priv->sh, priv->hws_cpool);
-		priv->hws_cpool = NULL;
-	}
-	mlx5_flow_quota_destroy(dev);
 	flow_hw_destroy_send_to_kernel_action(priv);
 	flow_hw_cleanup_ctrl_fdb_tables(dev);
 	flow_hw_free_vport_actions(priv);
+	if (priv->hw_def_miss) {
+		mlx5dr_action_destroy(priv->hw_def_miss);
+		priv->hw_def_miss = NULL;
+	}
+	flow_hw_cleanup_tx_repr_tagging(dev);
 	for (i = 0; i < MLX5_HW_ACTION_FLAG_MAX; i++) {
-		if (priv->hw_drop[i])
+		if (priv->hw_drop[i]) {
 			mlx5dr_action_destroy(priv->hw_drop[i]);
-		if (priv->hw_tag[i])
+			priv->hw_drop[i] = NULL;
+		}
+		if (priv->hw_tag[i]) {
 			mlx5dr_action_destroy(priv->hw_tag[i]);
+			priv->hw_tag[i] = NULL;
+		}
 	}
-	if (priv->hw_def_miss)
-		mlx5dr_action_destroy(priv->hw_def_miss);
-	flow_hw_destroy_vlan(dev);
-	if (dr_ctx)
+	mlx5_flow_meter_uninit(dev);
+	mlx5_flow_quota_destroy(dev);
+	flow_hw_cleanup_ctrl_rx_tables(dev);
+	if (dr_ctx) {
 		claim_zero(mlx5dr_context_close(dr_ctx));
+		priv->dr_ctx = NULL;
+	}
+	if (priv->shared_host) {
+		struct mlx5_priv *host_priv = priv->shared_host->data->dev_private;
+
+		__atomic_fetch_sub(&host_priv->shared_refcnt, 1, __ATOMIC_RELAXED);
+		priv->shared_host = NULL;
+	}
 	if (priv->hw_q) {
 		for (i = 0; i < nb_q_updated; i++) {
 			rte_ring_free(priv->hw_q[i].indir_iq);
@@ -9746,14 +9764,11 @@ flow_hw_configure(struct rte_eth_dev *dev,
 		mlx5_ipool_destroy(priv->acts_ipool);
 		priv->acts_ipool = NULL;
 	}
-	if (_queue_attr)
-		mlx5_free(_queue_attr);
-	if (priv->shared_host) {
-		__atomic_fetch_sub(&host_priv->shared_refcnt, 1, __ATOMIC_RELAXED);
-		priv->shared_host = NULL;
-	}
 	mlx5_free(priv->hw_attr);
 	priv->hw_attr = NULL;
+	priv->nb_queue = 0;
+	if (_queue_attr)
+		mlx5_free(_queue_attr);
 	/* Do not overwrite the internal errno information. */
 	if (ret)
 		return ret;
-- 
2.39.2