diff --git a/drivers/net/mlx5/linux/mlx5_os.c b/drivers/net/mlx5/linux/mlx5_os.c
index ef7ccba5de..534a56a555 100644
--- a/drivers/net/mlx5/linux/mlx5_os.c
+++ b/drivers/net/mlx5/linux/mlx5_os.c
@@ -1632,7 +1632,10 @@ mlx5_dev_spawn(struct rte_device *dpdk_dev,
 		priv->obj_ops.txq_obj_new = mlx5_os_txq_obj_new;
 		priv->obj_ops.txq_obj_release = mlx5_os_txq_obj_release;
 		mlx5_queue_counter_id_prepare(eth_dev);
-
+		priv->obj_ops.lb_dummy_queue_create =
+					mlx5_rxq_ibv_obj_dummy_lb_create;
+		priv->obj_ops.lb_dummy_queue_release =
+					mlx5_rxq_ibv_obj_dummy_lb_release;
 	} else {
 		priv->obj_ops = ibv_obj_ops;
 	}
diff --git a/drivers/net/mlx5/linux/mlx5_verbs.c b/drivers/net/mlx5/linux/mlx5_verbs.c
index 0b0759f33f..d4fa202ac4 100644
--- a/drivers/net/mlx5/linux/mlx5_verbs.c
+++ b/drivers/net/mlx5/linux/mlx5_verbs.c
@@ -1055,6 +1055,125 @@ mlx5_txq_ibv_obj_new(struct rte_eth_dev *dev, uint16_t idx)
 	return -rte_errno;
 }
 
+/*
+ * Create the dummy QP with minimal resources for loopback.
+ *
+ * @param dev
+ *   Pointer to Ethernet device.
+ *
+ * @return
+ *   0 on success, a negative errno value otherwise and rte_errno is set.
+ */
+int
+mlx5_rxq_ibv_obj_dummy_lb_create(struct rte_eth_dev *dev)
+{
+#if defined(HAVE_IBV_DEVICE_TUNNEL_SUPPORT) && defined(HAVE_IBV_FLOW_DV_SUPPORT)
+	struct mlx5_priv *priv = dev->data->dev_private;
+	struct mlx5_dev_ctx_shared *sh = priv->sh;
+	struct ibv_context *ctx = sh->ctx;
+	struct mlx5dv_qp_init_attr qp_init_attr = {0};
+	struct {
+		struct ibv_cq_init_attr_ex ibv;
+		struct mlx5dv_cq_init_attr mlx5;
+	} cq_attr = {{0}};
+
+	if (dev->data->dev_conf.lpbk_mode) {
+		/* Allow packet sent from NIC loop back w/o source MAC check. */
+		qp_init_attr.comp_mask |=
+				MLX5DV_QP_INIT_ATTR_MASK_QP_CREATE_FLAGS;
+		qp_init_attr.create_flags |=
+				MLX5DV_QP_CREATE_TIR_ALLOW_SELF_LOOPBACK_UC;
+	} else {
+		return 0;
+	}
+	/* Only need to check refcnt, 0 after "sh" is allocated. */
+	if (!!(__atomic_fetch_add(&sh->self_lb.refcnt, 1, __ATOMIC_RELAXED))) {
+		MLX5_ASSERT(sh->self_lb.ibv_cq && sh->self_lb.qp);
+		priv->lb_used = 1;
+		return 0;
+	}
+	cq_attr.ibv = (struct ibv_cq_init_attr_ex){
+		.cqe = 1,
+		.channel = NULL,
+		.comp_mask = 0,
+	};
+	cq_attr.mlx5 = (struct mlx5dv_cq_init_attr){
+		.comp_mask = 0,
+	};
+	/* Only CQ is needed, no WQ(RQ) is required in this case. */
+	sh->self_lb.ibv_cq = mlx5_glue->cq_ex_to_cq(mlx5_glue->dv_create_cq(ctx,
+							&cq_attr.ibv,
+							&cq_attr.mlx5));
+	if (!sh->self_lb.ibv_cq) {
+		DRV_LOG(ERR, "Port %u cannot allocate CQ for loopback.",
+			dev->data->port_id);
+		rte_errno = errno;
+		goto error;
+	}
+	sh->self_lb.qp = mlx5_glue->dv_create_qp(ctx,
+				&(struct ibv_qp_init_attr_ex){
+					.qp_type = IBV_QPT_RAW_PACKET,
+					.comp_mask = IBV_QP_INIT_ATTR_PD,
+					.pd = sh->pd,
+					.send_cq = sh->self_lb.ibv_cq,
+					.recv_cq = sh->self_lb.ibv_cq,
+					.cap.max_recv_wr = 1,
+				},
+				&qp_init_attr);
+	if (!sh->self_lb.qp) {
+		DRV_LOG(DEBUG, "Port %u cannot allocate QP for loopback.",
+			dev->data->port_id);
+		rte_errno = errno;
+		goto error;
+	}
+	priv->lb_used = 1;
+	return 0;
+error:
+	if (sh->self_lb.ibv_cq) {
+		claim_zero(mlx5_glue->destroy_cq(sh->self_lb.ibv_cq));
+		sh->self_lb.ibv_cq = NULL;
+	}
+	(void)__atomic_sub_fetch(&sh->self_lb.refcnt, 1, __ATOMIC_RELAXED);
+	return -rte_errno;
+#else
+	RTE_SET_USED(dev);
+	return 0;
+#endif
+}
+
+/*
+ * Release the dummy queue resources for loopback.
+ *
+ * @param dev
+ *   Pointer to Ethernet device.
+ */
+void
+mlx5_rxq_ibv_obj_dummy_lb_release(struct rte_eth_dev *dev)
+{
+#if defined(HAVE_IBV_DEVICE_TUNNEL_SUPPORT) && defined(HAVE_IBV_FLOW_DV_SUPPORT)
+	struct mlx5_priv *priv = dev->data->dev_private;
+	struct mlx5_dev_ctx_shared *sh = priv->sh;
+
+	if (!priv->lb_used)
+		return;
+	MLX5_ASSERT(__atomic_load_n(&sh->self_lb.refcnt, __ATOMIC_RELAXED));
+	if (!(__atomic_sub_fetch(&sh->self_lb.refcnt, 1, __ATOMIC_RELAXED))) {
+		if (sh->self_lb.qp) {
+			claim_zero(mlx5_glue->destroy_qp(sh->self_lb.qp));
+			sh->self_lb.qp = NULL;
+		}
+		if (sh->self_lb.ibv_cq) {
+			claim_zero(mlx5_glue->destroy_cq(sh->self_lb.ibv_cq));
+			sh->self_lb.ibv_cq = NULL;
+		}
+	}
+	priv->lb_used = 0;
+#else
+	RTE_SET_USED(dev);
+	return;
+#endif
+}
+
 /**
  * Release an Tx verbs queue object.
  *
@@ -1084,4 +1203,6 @@ struct mlx5_obj_ops ibv_obj_ops = {
 	.txq_obj_new = mlx5_txq_ibv_obj_new,
 	.txq_obj_modify = mlx5_ibv_modify_qp,
 	.txq_obj_release = mlx5_txq_ibv_obj_release,
+	.lb_dummy_queue_create = NULL,
+	.lb_dummy_queue_release = NULL,
 };
diff --git a/drivers/net/mlx5/linux/mlx5_verbs.h b/drivers/net/mlx5/linux/mlx5_verbs.h
index 76a79bf4f4..f7e8e2fe98 100644
--- a/drivers/net/mlx5/linux/mlx5_verbs.h
+++ b/drivers/net/mlx5/linux/mlx5_verbs.h
@@ -9,6 +9,8 @@
 
 int mlx5_txq_ibv_obj_new(struct rte_eth_dev *dev, uint16_t idx);
 void mlx5_txq_ibv_obj_release(struct mlx5_txq_obj *txq_obj);
+int mlx5_rxq_ibv_obj_dummy_lb_create(struct rte_eth_dev *dev);
+void mlx5_rxq_ibv_obj_dummy_lb_release(struct rte_eth_dev *dev);
 
 /* Verbs ops struct */
 extern const struct mlx5_mr_ops mlx5_mr_verbs_ops;
diff --git a/drivers/net/mlx5/mlx5.h b/drivers/net/mlx5/mlx5.h
index b8a29dd369..32b2817bf2 100644
--- a/drivers/net/mlx5/mlx5.h
+++ b/drivers/net/mlx5/mlx5.h
@@ -287,6 +287,13 @@ struct mlx5_drop {
 	struct mlx5_rxq_obj *rxq; /* Rx queue object. */
 };
 
+/* Loopback dummy queue resources required due to Verbs API. */
+struct mlx5_lb_ctx {
+	struct ibv_qp *qp; /* QP object. */
+	void *ibv_cq; /* Completion queue. */
+	uint16_t refcnt; /* Reference count for representors. */
+};
+
 #define MLX5_COUNTERS_PER_POOL 512
 #define MLX5_MAX_PENDING_QUERIES 4
 #define MLX5_CNT_CONTAINER_RESIZE 64
@@ -1128,6 +1135,7 @@ struct mlx5_dev_ctx_shared {
 	/* Meter management structure. */
 	struct mlx5_aso_ct_pools_mng *ct_mng;
 	/* Management data for ASO connection tracking. */
+	struct mlx5_lb_ctx self_lb; /* QP to enable self loopback for Devx. */
 	struct mlx5_dev_shared_port port[]; /* per device port data array. */
 };
 
@@ -1287,6 +1295,8 @@ struct mlx5_obj_ops {
 	int (*txq_obj_modify)(struct mlx5_txq_obj *obj,
 			      enum mlx5_txq_modify_type type, uint8_t dev_port);
 	void (*txq_obj_release)(struct mlx5_txq_obj *txq_obj);
+	int (*lb_dummy_queue_create)(struct rte_eth_dev *dev);
+	void (*lb_dummy_queue_release)(struct rte_eth_dev *dev);
 };
 
 #define MLX5_RSS_HASH_FIELDS_LEN RTE_DIM(mlx5_rss_hash_fields)
@@ -1316,6 +1326,7 @@ struct mlx5_priv {
 	unsigned int sampler_en:1; /* Whether support sampler. */
 	unsigned int mtr_en:1; /* Whether support meter. */
 	unsigned int mtr_reg_share:1; /* Whether support meter REG_C share. */
+	unsigned int lb_used:1; /* Loopback queue is referred to. */
 	uint16_t domain_id; /* Switch domain identifier. */
 	uint16_t vport_id; /* Associated VF vport index (if any). */
 	uint32_t vport_meta_tag; /* Used for vport index match ove VF LAG. */
diff --git a/drivers/net/mlx5/mlx5_devx.c b/drivers/net/mlx5/mlx5_devx.c
index 531a81d7fa..78b88f99b4 100644
--- a/drivers/net/mlx5/mlx5_devx.c
+++ b/drivers/net/mlx5/mlx5_devx.c
@@ -1188,4 +1188,6 @@ struct mlx5_obj_ops devx_obj_ops = {
 	.txq_obj_new = mlx5_txq_devx_obj_new,
 	.txq_obj_modify = mlx5_devx_modify_sq,
 	.txq_obj_release = mlx5_txq_devx_obj_release,
+	.lb_dummy_queue_create = NULL,
+	.lb_dummy_queue_release = NULL,
 };
diff --git a/drivers/net/mlx5/mlx5_trigger.c b/drivers/net/mlx5/mlx5_trigger.c
index 879d3171e9..ae7fcca229 100644
--- a/drivers/net/mlx5/mlx5_trigger.c
+++ b/drivers/net/mlx5/mlx5_trigger.c
@@ -1068,6 +1068,12 @@ mlx5_dev_start(struct rte_eth_dev *dev)
 			dev->data->port_id, strerror(rte_errno));
 		goto error;
 	}
+	if ((priv->config.devx && priv->config.dv_flow_en &&
+	    priv->config.dest_tir) && priv->obj_ops.lb_dummy_queue_create) {
+		ret = priv->obj_ops.lb_dummy_queue_create(dev);
+		if (ret)
+			goto error;
+	}
 	ret = mlx5_txq_start(dev);
 	if (ret) {
 		DRV_LOG(ERR, "port %u Tx queue allocation failed: %s",
@@ -1148,6 +1154,8 @@ mlx5_dev_start(struct rte_eth_dev *dev)
 	mlx5_traffic_disable(dev);
 	mlx5_txq_stop(dev);
 	mlx5_rxq_stop(dev);
+	if (priv->obj_ops.lb_dummy_queue_release)
+		priv->obj_ops.lb_dummy_queue_release(dev);
 	mlx5_txpp_stop(dev); /* Stop last. */
 	rte_errno = ret; /* Restore rte_errno. */
 	return -rte_errno;
@@ -1186,6 +1194,8 @@ mlx5_dev_stop(struct rte_eth_dev *dev)
 	priv->sh->port[priv->dev_port - 1].devx_ih_port_id = RTE_MAX_ETHPORTS;
 	mlx5_txq_stop(dev);
 	mlx5_rxq_stop(dev);
+	if (priv->obj_ops.lb_dummy_queue_release)
+		priv->obj_ops.lb_dummy_queue_release(dev);
 	mlx5_txpp_stop(dev);
 
 	return 0;