net/mlx5e: Support HTB offload

This commit adds support for HTB offload in the mlx5e driver.

Performance:

  NIC: Mellanox ConnectX-6 Dx
  CPU: Intel(R) Xeon(R) CPU E5-2680 v3 @ 2.50GHz (24 cores with HT)

  100 Gbit/s line rate, 500 UDP streams @ ~200 Mbit/s each
  48 traffic classes, flower used for steering
  No shaping (rate limits set to 4 Gbit/s per TC) - checking for max
  throughput.

  Baseline: 98.7 Gbps, 8.25 Mpps
  HTB: 6.7 Gbps, 0.56 Mpps
  HTB offload: 95.6 Gbps, 8.00 Mpps

Limitations:

1. 256 leaf nodes, 3 levels of depth.

2. Granularity for ceil is 1 Mbit/s. Rates are converted to weights, and
the bandwidth is split among the siblings according to these weights.
Other parameters for classes are not supported.

Ethtool statistics support for QoS SQs are also added. The counters are
called qos_txN_*, where N is the QoS queue number (starting from 0, the
numeration is separate from the normal SQs), and * is the counter name
(the counters are the same as for the normal SQs).

Signed-off-by: Maxim Mikityanskiy <maximmi@mellanox.com>
Reviewed-by: Tariq Toukan <tariqt@nvidia.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

1 files changed, 38 insertions, 9 deletions

diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en_tx.c b/drivers/net/ethernet/mellanox/mlx5/core/en_tx.c
index 74f233eece54..da6a358a8a10 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en_tx.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en_tx.c
@@ -106,28 +106,53 @@ return_txq:
 	return priv->port_ptp_tc2realtxq[up];
 }
 
+static int mlx5e_select_htb_queue(struct mlx5e_priv *priv, struct sk_buff *skb,
+				  u16 htb_maj_id)
+{
+	u16 classid;
+
+	if ((TC_H_MAJ(skb->priority) >> 16) == htb_maj_id)
+		classid = TC_H_MIN(skb->priority);
+	else
+		classid = READ_ONCE(priv->htb.defcls);
+
+	if (!classid)
+		return 0;
+
+	return mlx5e_get_txq_by_classid(priv, classid);
+}
+
 u16 mlx5e_select_queue(struct net_device *dev, struct sk_buff *skb,
 		       struct net_device *sb_dev)
 {
 	struct mlx5e_priv *priv = netdev_priv(dev);
+	int num_tc_x_num_ch;
 	int txq_ix;
 	int up = 0;
 	int ch_ix;
 
-	if (unlikely(priv->channels.port_ptp)) {
-		int num_tc_x_num_ch;
+	/* Sync with mlx5e_update_num_tc_x_num_ch - avoid refetching. */
+	num_tc_x_num_ch = READ_ONCE(priv->num_tc_x_num_ch);
+	if (unlikely(dev->real_num_tx_queues > num_tc_x_num_ch)) {
+		/* Order maj_id before defcls - pairs with mlx5e_htb_root_add. */
+		u16 htb_maj_id = smp_load_acquire(&priv->htb.maj_id);
 
-		if (unlikely(skb_shinfo(skb)->tx_flags & SKBTX_HW_TSTAMP) &&
-		    mlx5e_use_ptpsq(skb))
-			return mlx5e_select_ptpsq(dev, skb);
+		if (unlikely(htb_maj_id)) {
+			txq_ix = mlx5e_select_htb_queue(priv, skb, htb_maj_id);
+			if (txq_ix > 0)
+				return txq_ix;
+		}
 
-		/* Sync with mlx5e_update_num_tc_x_num_ch - avoid refetching. */
-		num_tc_x_num_ch = READ_ONCE(priv->num_tc_x_num_ch);
+		if (unlikely(priv->channels.port_ptp))
+			if (unlikely(skb_shinfo(skb)->tx_flags & SKBTX_HW_TSTAMP) &&
+			    mlx5e_use_ptpsq(skb))
+				return mlx5e_select_ptpsq(dev, skb);
 
 		txq_ix = netdev_pick_tx(dev, skb, NULL);
-		/* Fix netdev_pick_tx() not to choose ptp_channel txqs.
+		/* Fix netdev_pick_tx() not to choose ptp_channel and HTB txqs.
 		 * If they are selected, switch to regular queues.
-		 * Driver to select these queues only at mlx5e_select_ptpsq().
+		 * Driver to select these queues only at mlx5e_select_ptpsq()
+		 * and mlx5e_select_htb_queue().
 		 */
 		if (unlikely(txq_ix >= num_tc_x_num_ch))
 			txq_ix %= num_tc_x_num_ch;
@@ -702,6 +727,10 @@ netdev_tx_t mlx5e_xmit(struct sk_buff *skb, struct net_device *dev)
 	u16 pi;
 
 	sq = priv->txq2sq[skb_get_queue_mapping(skb)];
+	if (unlikely(!sq)) {
+		dev_kfree_skb_any(skb);
+		return NETDEV_TX_OK;
+	}
 
 	/* May send SKBs and WQEs. */
 	if (unlikely(!mlx5e_accel_tx_begin(dev, sq, skb, &accel)))