net/mlx5e: Don't trigger IRQ multiple times on XSK wakeup to avoid WQ overruns

XSK wakeup function triggers NAPI by posting a NOP WQE to a special XSK
ICOSQ. When the application floods the driver with wakeup requests by
calling sendto() in a certain pattern that ends up in mlx5e_trigger_irq,
the XSK ICOSQ may overflow.

Multiple NOPs are not required and won't accelerate the process, so
avoid posting a second NOP if there is one already on the way. This way
we also avoid increasing the queue size (which might not help anyway).

Fixes: db05815 ("net/mlx5e: Add XSK zero-copy support")
Signed-off-by: Maxim Mikityanskiy <[email protected]>
Reviewed-by: Tariq Toukan <[email protected]>
Signed-off-by: Saeed Mahameed <[email protected]>

Author: Maxim Mikityanskiy

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

@@ -367,6 +367,7 @@ enum {
 	MLX5E_SQ_STATE_AM,
 	MLX5E_SQ_STATE_TLS,
 	MLX5E_SQ_STATE_VLAN_NEED_L2_INLINE,
+	MLX5E_SQ_STATE_PENDING_XSK_TX,
 };
 
 struct mlx5e_sq_wqe_info {
@@ -960,7 +961,7 @@ void mlx5e_page_release_dynamic(struct mlx5e_rq *rq,
 void mlx5e_handle_rx_cqe(struct mlx5e_rq *rq, struct mlx5_cqe64 *cqe);
 void mlx5e_handle_rx_cqe_mpwrq(struct mlx5e_rq *rq, struct mlx5_cqe64 *cqe);
 bool mlx5e_post_rx_wqes(struct mlx5e_rq *rq);
-void mlx5e_poll_ico_cq(struct mlx5e_cq *cq);
+int mlx5e_poll_ico_cq(struct mlx5e_cq *cq);
 bool mlx5e_post_rx_mpwqes(struct mlx5e_rq *rq);
 void mlx5e_dealloc_rx_wqe(struct mlx5e_rq *rq, u16 ix);
 void mlx5e_dealloc_rx_mpwqe(struct mlx5e_rq *rq, u16 ix);

drivers/net/ethernet/mellanox/mlx5/core/en/xsk/tx.c

@@ -33,6 +33,9 @@ int mlx5e_xsk_wakeup(struct net_device *dev, u32 qid, u32 flags)
 		if (unlikely(!test_bit(MLX5E_SQ_STATE_ENABLED, &c->xskicosq.state)))
 			return 0;
 
+		if (test_and_set_bit(MLX5E_SQ_STATE_PENDING_XSK_TX, &c->xskicosq.state))
+			return 0;
+
 		spin_lock(&c->xskicosq_lock);
 		mlx5e_trigger_irq(&c->xskicosq);
 		spin_unlock(&c->xskicosq_lock);

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

@@ -589,19 +589,19 @@ bool mlx5e_post_rx_wqes(struct mlx5e_rq *rq)
 	return !!err;
 }
 
-void mlx5e_poll_ico_cq(struct mlx5e_cq *cq)
+int mlx5e_poll_ico_cq(struct mlx5e_cq *cq)
 {
 	struct mlx5e_icosq *sq = container_of(cq, struct mlx5e_icosq, cq);
 	struct mlx5_cqe64 *cqe;
 	u16 sqcc;
 	int i;
 
 	if (unlikely(!test_bit(MLX5E_SQ_STATE_ENABLED, &sq->state)))
-		return;
+		return 0;
 
 	cqe = mlx5_cqwq_get_cqe(&cq->wq);
 	if (likely(!cqe))
-		return;
+		return 0;
 
 	/* sq->cc must be updated only after mlx5_cqwq_update_db_record(),
 	 * otherwise a cq overrun may occur
@@ -650,6 +650,8 @@ void mlx5e_poll_ico_cq(struct mlx5e_cq *cq)
 	sq->cc = sqcc;
 
 	mlx5_cqwq_update_db_record(&cq->wq);
+
+	return i;
 }
 
 bool mlx5e_post_rx_mpwqes(struct mlx5e_rq *rq)

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

@@ -152,7 +152,11 @@ int mlx5e_napi_poll(struct napi_struct *napi, int budget)
 				mlx5e_post_rx_wqes,
 				rq);
 	if (xsk_open) {
-		mlx5e_poll_ico_cq(&c->xskicosq.cq);
+		if (mlx5e_poll_ico_cq(&c->xskicosq.cq))
+			/* Don't clear the flag if nothing was polled to prevent
+			 * queueing more WQEs and overflowing XSKICOSQ.
+			 */
+			clear_bit(MLX5E_SQ_STATE_PENDING_XSK_TX, &c->xskicosq.state);
 		busy |= mlx5e_poll_xdpsq_cq(&xsksq->cq);
 		busy_xsk |= mlx5e_napi_xsk_post(xsksq, xskrq);
 	}