From: Viacheslav Ovsiienko <viacheslavo@mellanox.com>
Date: Wed, 7 Aug 2019 12:58:10 +0000 (+0000)
Subject: net/mlx5: fix completion request for multi-segment
X-Git-Url: http://git.droids-corp.org/?a=commitdiff_plain;h=4dec9c79c2d76e08b3951a91137bb8622a465675;p=dpdk.git

net/mlx5: fix completion request for multi-segment

The copying of sent mbufs pointers might be deferred to the end of
tx_burst() routine to be copied in one call of rte_memcpy.
For the multi segment packets this optimization is not applicable,
because number of packets does not match with number of mbufs and
we do not have linear array of pointers in pkts parameter.

The completion request generating routine wrongly took into account
the inconsistent (for multi-segment packets) deferred pointer copying.

Fixes: 5a93e173b874 ("net/mlx5: fix Tx completion request generation")

Signed-off-by: Viacheslav Ovsiienko <viacheslavo@mellanox.com>
---

diff --git a/drivers/net/mlx5/mlx5_rxtx.c b/drivers/net/mlx5/mlx5_rxtx.c
index 3e51bb00b8..4c01187faa 100644
--- a/drivers/net/mlx5/mlx5_rxtx.c
+++ b/drivers/net/mlx5/mlx5_rxtx.c
@@ -2130,6 +2130,9 @@ mlx5_tx_handle_completion(struct mlx5_txq_data *restrict txq,
  *   Pointer to TX queue structure.
  * @param loc
  *   Pointer to burst routine local context.
+ * @param multi,
+ *   Routine is called from multi-segment sending loop,
+ *   do not correct the elts_head according to the pkts_copy.
  * @param olx
  *   Configured Tx offloads mask. It is fully defined at
  *   compile time and may be used for optimization.
@@ -2137,12 +2140,14 @@ mlx5_tx_handle_completion(struct mlx5_txq_data *restrict txq,
 static __rte_always_inline void
 mlx5_tx_request_completion(struct mlx5_txq_data *restrict txq,
 			   struct mlx5_txq_local *restrict loc,
+			   bool multi,
 			   unsigned int olx)
 {
 	uint16_t head = txq->elts_head;
 	unsigned int part;
 
-	part = MLX5_TXOFF_CONFIG(INLINE) ? 0 : loc->pkts_sent - loc->pkts_copy;
+	part = (MLX5_TXOFF_CONFIG(INLINE) || multi) ?
+	       0 : loc->pkts_sent - loc->pkts_copy;
 	head += part;
 	if ((uint16_t)(head - txq->elts_comp) >= MLX5_TX_COMP_THRESH ||
 	     (MLX5_TXOFF_CONFIG(INLINE) &&
@@ -3091,7 +3096,7 @@ mlx5_tx_packet_multi_tso(struct mlx5_txq_data *restrict txq,
 	txq->wqe_ci += (ds + 3) / 4;
 	loc->wqe_free -= (ds + 3) / 4;
 	/* Request CQE generation if limits are reached. */
-	mlx5_tx_request_completion(txq, loc, olx);
+	mlx5_tx_request_completion(txq, loc, true, olx);
 	return MLX5_TXCMP_CODE_MULTI;
 }
 
@@ -3201,7 +3206,7 @@ mlx5_tx_packet_multi_send(struct mlx5_txq_data *restrict txq,
 	txq->wqe_ci += (ds + 3) / 4;
 	loc->wqe_free -= (ds + 3) / 4;
 	/* Request CQE generation if limits are reached. */
-	mlx5_tx_request_completion(txq, loc, olx);
+	mlx5_tx_request_completion(txq, loc, true, olx);
 	return MLX5_TXCMP_CODE_MULTI;
 }
 
@@ -3359,7 +3364,7 @@ do_align:
 	txq->wqe_ci += (ds + 3) / 4;
 	loc->wqe_free -= (ds + 3) / 4;
 	/* Request CQE generation if limits are reached. */
-	mlx5_tx_request_completion(txq, loc, olx);
+	mlx5_tx_request_completion(txq, loc, true, olx);
 	return MLX5_TXCMP_CODE_MULTI;
 }
 
@@ -3570,7 +3575,7 @@ mlx5_tx_burst_tso(struct mlx5_txq_data *restrict txq,
 		++loc->pkts_sent;
 		--pkts_n;
 		/* Request CQE generation if limits are reached. */
-		mlx5_tx_request_completion(txq, loc, olx);
+		mlx5_tx_request_completion(txq, loc, false, olx);
 		if (unlikely(!pkts_n || !loc->elts_free || !loc->wqe_free))
 			return MLX5_TXCMP_CODE_EXIT;
 		loc->mbuf = *pkts++;
@@ -3728,7 +3733,7 @@ mlx5_tx_sdone_empw(struct mlx5_txq_data *restrict txq,
 	txq->wqe_ci += (ds + 3) / 4;
 	loc->wqe_free -= (ds + 3) / 4;
 	/* Request CQE generation if limits are reached. */
-	mlx5_tx_request_completion(txq, loc, olx);
+	mlx5_tx_request_completion(txq, loc, false, olx);
 }
 
 /*
@@ -3772,7 +3777,7 @@ mlx5_tx_idone_empw(struct mlx5_txq_data *restrict txq,
 	txq->wqe_ci += (len + 3) / 4;
 	loc->wqe_free -= (len + 3) / 4;
 	/* Request CQE generation if limits are reached. */
-	mlx5_tx_request_completion(txq, loc, olx);
+	mlx5_tx_request_completion(txq, loc, false, olx);
 }
 
 /**
@@ -3965,7 +3970,7 @@ next_empw:
 		loc->wqe_free -= (2 + part + 3) / 4;
 		pkts_n -= part;
 		/* Request CQE generation if limits are reached. */
-		mlx5_tx_request_completion(txq, loc, olx);
+		mlx5_tx_request_completion(txq, loc, false, olx);
 		if (unlikely(!pkts_n || !loc->elts_free || !loc->wqe_free))
 			return MLX5_TXCMP_CODE_EXIT;
 		loc->mbuf = *pkts++;
@@ -4440,7 +4445,7 @@ mlx5_tx_burst_single_send(struct mlx5_txq_data *restrict txq,
 		++loc->pkts_sent;
 		--pkts_n;
 		/* Request CQE generation if limits are reached. */
-		mlx5_tx_request_completion(txq, loc, olx);
+		mlx5_tx_request_completion(txq, loc, false, olx);
 		if (unlikely(!pkts_n || !loc->elts_free || !loc->wqe_free))
 			return MLX5_TXCMP_CODE_EXIT;
 		loc->mbuf = *pkts++;