net/mlx5e: Support NETIF_F_SG

When NETIF_F_SG is set, each send WQE may have a different size since
each skb can have different number of fragments as of LSO header etc.

This implies that a given WQE may wrap around the send queue, i.e begin
at its end and continue at its start. While it is legal by the device spec,
we preferred a solution that avoids it - when building of current WQE is
done, if the next WQE may wrap around the send queue, fill the send queue
with NOPs WQEs till its end, so that the next WQE will begin at send queue
start.

NOP WQE for itself cannot wrap around the send queue since it is of
minimal size - 64 bytes, and all send WQEs are a multiple of that size.

Signed-off-by: Achiad Shochat <achiad@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

1 files changed, 33 insertions, 2 deletions

diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en_tx.c b/drivers/net/ethernet/mellanox/mlx5/core/en_tx.c
index 3cfd2bcb3c8b..bac268a670f4 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en_tx.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en_tx.c
@@ -34,6 +34,33 @@
 #include <linux/if_vlan.h>
 #include "en.h"
 
+#define MLX5E_SQ_NOPS_ROOM  MLX5_SEND_WQE_MAX_WQEBBS
+#define MLX5E_SQ_STOP_ROOM (MLX5_SEND_WQE_MAX_WQEBBS +\
+			    MLX5E_SQ_NOPS_ROOM)
+
+void mlx5e_send_nop(struct mlx5e_sq *sq, bool notify_hw)
+{
+	struct mlx5_wq_cyc                *wq  = &sq->wq;
+
+	u16 pi = sq->pc & wq->sz_m1;
+	struct mlx5e_tx_wqe              *wqe  = mlx5_wq_cyc_get_wqe(wq, pi);
+
+	struct mlx5_wqe_ctrl_seg         *cseg = &wqe->ctrl;
+
+	memset(cseg, 0, sizeof(*cseg));
+
+	cseg->opmod_idx_opcode = cpu_to_be32((sq->pc << 8) | MLX5_OPCODE_NOP);
+	cseg->qpn_ds           = cpu_to_be32((sq->sqn << 8) | 0x01);
+
+	sq->skb[pi] = NULL;
+	sq->pc++;
+
+	if (notify_hw) {
+		cseg->fm_ce_se = MLX5_WQE_CTRL_CQ_UPDATE;
+		mlx5e_tx_notify_hw(sq, wqe);
+	}
+}
+
 static void mlx5e_dma_pop_last_pushed(struct mlx5e_sq *sq, dma_addr_t *addr,
 				      u32 *size)
 {
@@ -196,7 +223,7 @@ static netdev_tx_t mlx5e_sq_xmit(struct mlx5e_sq *sq, struct sk_buff *skb)
 
 	netdev_tx_sent_queue(sq->txq, MLX5E_TX_SKB_CB(skb)->num_bytes);
 
-	if (unlikely(!mlx5e_sq_has_room_for(sq, MLX5_SEND_WQE_MAX_WQEBBS))) {
+	if (unlikely(!mlx5e_sq_has_room_for(sq, MLX5E_SQ_STOP_ROOM))) {
 		netif_tx_stop_queue(sq->txq);
 		sq->stats.stopped++;
 	}
@@ -204,6 +231,10 @@ static netdev_tx_t mlx5e_sq_xmit(struct mlx5e_sq *sq, struct sk_buff *skb)
 	if (!skb->xmit_more || netif_xmit_stopped(sq->txq))
 		mlx5e_tx_notify_hw(sq, wqe);
 
+	/* fill sq edge with nops to avoid wqe wrap around */
+	while ((sq->pc & wq->sz_m1) > sq->edge)
+		mlx5e_send_nop(sq, false);
+
 	sq->stats.packets++;
 	return NETDEV_TX_OK;
 
@@ -311,7 +342,7 @@ free_skb:
 	netdev_tx_completed_queue(sq->txq, npkts, nbytes);
 
 	if (netif_tx_queue_stopped(sq->txq) &&
-	    mlx5e_sq_has_room_for(sq, MLX5_SEND_WQE_MAX_WQEBBS) &&
+	    mlx5e_sq_has_room_for(sq, MLX5E_SQ_STOP_ROOM) &&
 	    likely(test_bit(MLX5E_SQ_STATE_WAKE_TXQ_ENABLE, &sq->state))) {
 				netif_tx_wake_queue(sq->txq);
 				sq->stats.wake++;