Merge branch 'mlx4-dynamic-tc-tx-queues'

Tariq Toukan says:

====================
mlx4_en dynamic TC tx queues

This patchset from Inbar aligns the number of TX queues
to the actual need, according to the TC configuration.

Series generated against net-next commit:
2ee87db3a287 Merge branch 'nfp-get_phys_port_name-for-representors-and-SR-IOV-reorder'
====================

Signed-off-by: David S. Miller <[e-mail]>

drivers/net/ethernet/mellanox/mlx4/en_dcb_nl.c

···
 		priv->flags &= ~MLX4_EN_FLAG_DCB_ENABLED;
 	}
 
-	if (mlx4_en_setup_tc(dev, num_tcs))
+	if (mlx4_en_alloc_tx_queue_per_tc(dev, num_tcs))
 		return 1;
 
 	return 0;
···
 	int has_ets_tc = 0;
 
 	for (i = 0; i < IEEE_8021QAZ_MAX_TCS; i++) {
-		if (ets->prio_tc[i] >= MLX4_EN_NUM_UP) {
+		if (ets->prio_tc[i] >= MLX4_EN_NUM_UP_HIGH) {
 			en_err(priv, "Bad priority in UP <=> TC mapping. TC: %d, UP: %d\n",
 					i, ets->prio_tc[i]);
 			return -EINVAL;
···
 			goto err;
 		if (mlx4_en_dcbnl_ieee_setpfc(dev, &pfc))
 			goto err;
-		if (mlx4_en_setup_tc(dev, 0))
+		if (mlx4_en_alloc_tx_queue_per_tc(dev, 0))
 			goto err;
 	}
 

drivers/net/ethernet/mellanox/mlx4/en_ethtool.c

···
 	channel->max_tx = MLX4_EN_MAX_TX_RING_P_UP;
 
 	channel->rx_count = priv->rx_ring_num;
-	channel->tx_count = priv->tx_ring_num[TX] / MLX4_EN_NUM_UP;
+	channel->tx_count = priv->tx_ring_num[TX] /
+			    priv->prof->num_up;
 }
 
 static int mlx4_en_set_channels(struct net_device *dev,
···
 	int port_up = 0;
 	int xdp_count;
 	int err = 0;
+	u8 up;
 
 	if (!channel->tx_count || !channel->rx_count)
 		return -EINVAL;
···
 
 	mutex_lock(&mdev->state_lock);
 	xdp_count = priv->tx_ring_num[TX_XDP] ? channel->rx_count : 0;
-	if (channel->tx_count * MLX4_EN_NUM_UP + xdp_count > MAX_TX_RINGS) {
+	if (channel->tx_count * priv->prof->num_up + xdp_count >
+	    MAX_TX_RINGS) {
 		err = -EINVAL;
 		en_err(priv,
 		       "Total number of TX and XDP rings (%d) exceeds the maximum supported (%d)\n",
-		       channel->tx_count * MLX4_EN_NUM_UP + xdp_count,
+		       channel->tx_count * priv->prof->num_up  + xdp_count,
 		       MAX_TX_RINGS);
 		goto out;
 	}
 
 	memcpy(&new_prof, priv->prof, sizeof(struct mlx4_en_port_profile));
 	new_prof.num_tx_rings_p_up = channel->tx_count;
-	new_prof.tx_ring_num[TX] = channel->tx_count * MLX4_EN_NUM_UP;
+	new_prof.tx_ring_num[TX] = channel->tx_count * priv->prof->num_up;
 	new_prof.tx_ring_num[TX_XDP] = xdp_count;
 	new_prof.rx_ring_num = channel->rx_count;
 
···
 
 	mlx4_en_safe_replace_resources(priv, tmp);
 
-	netif_set_real_num_tx_queues(dev, priv->tx_ring_num[TX]);
 	netif_set_real_num_rx_queues(dev, priv->rx_ring_num);
 
-	if (netdev_get_num_tc(dev))
-		mlx4_en_setup_tc(dev, MLX4_EN_NUM_UP);
+	up = (priv->prof->num_up == MLX4_EN_NUM_UP_LOW) ?
+				    0 : priv->prof->num_up;
+	mlx4_en_setup_tc(dev, up);
 
 	en_warn(priv, "Using %d TX rings\n", priv->tx_ring_num[TX]);
 	en_warn(priv, "Using %d RX rings\n", priv->rx_ring_num);

drivers/net/ethernet/mellanox/mlx4/en_main.c

···
 		params->prof[i].tx_ppp = pfctx;
 		params->prof[i].tx_ring_size = MLX4_EN_DEF_TX_RING_SIZE;
 		params->prof[i].rx_ring_size = MLX4_EN_DEF_RX_RING_SIZE;
+		params->prof[i].num_up = MLX4_EN_NUM_UP_LOW;
+		params->prof[i].num_tx_rings_p_up = params->num_tx_rings_p_up;
 		params->prof[i].tx_ring_num[TX] = params->num_tx_rings_p_up *
-			MLX4_EN_NUM_UP;
+			params->prof[i].num_up;
 		params->prof[i].rss_rings = 0;
 		params->prof[i].inline_thold = inline_thold;
 	}

drivers/net/ethernet/mellanox/mlx4/en_netdev.c

···
 	int i;
 	unsigned int offset = 0;
 
-	if (up && up != MLX4_EN_NUM_UP)
+	if (up && up != MLX4_EN_NUM_UP_HIGH)
 		return -EINVAL;
 
 	netdev_set_num_tc(dev, up);
-
+	netif_set_real_num_tx_queues(dev, priv->tx_ring_num[TX]);
 	/* Partition Tx queues evenly amongst UP's */
 	for (i = 0; i < up; i++) {
 		netdev_set_tc_queue(dev, i, priv->num_tx_rings_p_up, offset);
···
 	return 0;
 }
 
+int mlx4_en_alloc_tx_queue_per_tc(struct net_device *dev, u8 tc)
+{
+	struct mlx4_en_priv *priv = netdev_priv(dev);
+	struct mlx4_en_dev *mdev = priv->mdev;
+	struct mlx4_en_port_profile new_prof;
+	struct mlx4_en_priv *tmp;
+	int port_up = 0;
+	int err = 0;
+
+	tmp = kzalloc(sizeof(*tmp), GFP_KERNEL);
+	if (!tmp)
+		return -ENOMEM;
+
+	mutex_lock(&mdev->state_lock);
+	memcpy(&new_prof, priv->prof, sizeof(struct mlx4_en_port_profile));
+	new_prof.num_up = (tc == 0) ? MLX4_EN_NUM_UP_LOW :
+				      MLX4_EN_NUM_UP_HIGH;
+	new_prof.tx_ring_num[TX] = new_prof.num_tx_rings_p_up *
+				   new_prof.num_up;
+	err = mlx4_en_try_alloc_resources(priv, tmp, &new_prof, true);
+	if (err)
+		goto out;
+
+	if (priv->port_up) {
+		port_up = 1;
+		mlx4_en_stop_port(dev, 1);
+	}
+
+	mlx4_en_safe_replace_resources(priv, tmp);
+	if (port_up) {
+		err = mlx4_en_start_port(dev);
+		if (err) {
+			en_err(priv, "Failed starting port for setup TC\n");
+			goto out;
+		}
+	}
+
+	err = mlx4_en_setup_tc(dev, tc);
+out:
+	mutex_unlock(&mdev->state_lock);
+	kfree(tmp);
+	return err;
+}
+
 static int __mlx4_en_setup_tc(struct net_device *dev, u32 handle,
 			      u32 chain_index, __be16 proto,
 			      struct tc_to_netdev *tc)
···
 	if (tc->type != TC_SETUP_MQPRIO)
 		return -EINVAL;
 
+	if (tc->mqprio->num_tc && tc->mqprio->num_tc != MLX4_EN_NUM_UP_HIGH)
+		return -EINVAL;
+
 	tc->mqprio->hw = TC_MQPRIO_HW_OFFLOAD_TCS;
 
-	return mlx4_en_setup_tc(dev, tc->mqprio->num_tc);
+	return mlx4_en_alloc_tx_queue_per_tc(dev, tc->mqprio->num_tc);
 }
 
 #ifdef CONFIG_RFS_ACCEL
···
 
 	memcpy(&dst->hwtstamp_config, &prof->hwtstamp_config,
 	       sizeof(dst->hwtstamp_config));
-	dst->num_tx_rings_p_up = src->mdev->profile.num_tx_rings_p_up;
+	dst->num_tx_rings_p_up = prof->num_tx_rings_p_up;
 	dst->rx_ring_num = prof->rx_ring_num;
 	dst->flags = prof->flags;
 	dst->mdev = src->mdev;
···
 		dst->tx_ring[t] = src->tx_ring[t];
 		dst->tx_cq[t] = src->tx_cq[t];
 	}
+	dst->num_tx_rings_p_up = src->num_tx_rings_p_up;
 	dst->rx_ring_num = src->rx_ring_num;
 	memcpy(dst->prof, src->prof, sizeof(struct mlx4_en_port_profile));
 }
···
 	if (priv->tx_ring_num[TX] + xdp_ring_num > MAX_TX_RINGS) {
 		tx_changed = 1;
 		new_prof.tx_ring_num[TX] =
-			MAX_TX_RINGS - ALIGN(xdp_ring_num, MLX4_EN_NUM_UP);
+			MAX_TX_RINGS - ALIGN(xdp_ring_num, priv->prof->num_up);
 		en_warn(priv, "Reducing the number of TX rings, to not exceed the max total rings number.\n");
 	}
 
···
 		priv->flags |= MLX4_EN_DCB_ENABLED;
 		priv->cee_config.pfc_state = false;
 
-		for (i = 0; i < MLX4_EN_NUM_UP; i++)
+		for (i = 0; i < MLX4_EN_NUM_UP_HIGH; i++)
 			priv->cee_config.dcb_pfc[i] = pfc_disabled;
 
 		if (mdev->dev->caps.flags2 & MLX4_DEV_CAP_FLAG2_ETS_CFG) {

drivers/net/ethernet/mellanox/mlx4/en_resources.c

···
 	context->local_qpn = cpu_to_be32(qpn);
 	context->pri_path.ackto = 1 & 0x07;
 	context->pri_path.sched_queue = 0x83 | (priv->port - 1) << 6;
-	if (user_prio >= 0) {
+	/* force user priority per tx ring */
+	if (user_prio >= 0 && priv->prof->num_up == MLX4_EN_NUM_UP_HIGH) {
 		context->pri_path.sched_queue |= user_prio << 3;
 		context->pri_path.feup = MLX4_FEUP_FORCE_ETH_UP;
 	}

drivers/net/ethernet/mellanox/mlx4/en_tx.c

···
 {
 	struct mlx4_en_priv *priv = netdev_priv(dev);
 	u16 rings_p_up = priv->num_tx_rings_p_up;
-	u8 up = 0;
 
 	if (netdev_get_num_tc(dev))
 		return skb_tx_hash(dev, skb);
 
-	if (skb_vlan_tag_present(skb))
-		up = skb_vlan_tag_get(skb) >> VLAN_PRIO_SHIFT;
-
-	return fallback(dev, skb) % rings_p_up + up * rings_p_up;
+	return fallback(dev, skb) % rings_p_up;
 }
 
 static void mlx4_bf_copy(void __iomem *dst, const void *src,

drivers/net/ethernet/mellanox/mlx4/mlx4_en.h

···
 #define MLX4_EN_SMALL_PKT_SIZE		64
 #define MLX4_EN_MIN_TX_RING_P_UP	1
 #define MLX4_EN_MAX_TX_RING_P_UP	32
-#define MLX4_EN_NUM_UP			8
+#define MLX4_EN_NUM_UP_LOW		1
+#define MLX4_EN_NUM_UP_HIGH		8
 #define MLX4_EN_DEF_RX_RING_SIZE  	1024
 #define MLX4_EN_DEF_TX_RING_SIZE	MLX4_EN_DEF_RX_RING_SIZE
 #define MAX_TX_RINGS			(MLX4_EN_MAX_TX_RING_P_UP * \
-					 MLX4_EN_NUM_UP)
+					 MLX4_EN_NUM_UP_HIGH)
 
 #define MLX4_EN_DEFAULT_TX_WORK		256
 
···
 	u8 rx_ppp;
 	u8 tx_pause;
 	u8 tx_ppp;
+	u8 num_up;
 	int rss_rings;
 	int inline_thold;
 	struct hwtstamp_config hwtstamp_config;
···
 
 struct mlx4_en_cee_config {
 	bool	pfc_state;
-	enum	dcb_pfc_type dcb_pfc[MLX4_EN_NUM_UP];
+	enum	dcb_pfc_type dcb_pfc[MLX4_EN_NUM_UP_HIGH];
 };
 #endif
 
···
 #endif
 
 int mlx4_en_setup_tc(struct net_device *dev, u8 up);
+int mlx4_en_alloc_tx_queue_per_tc(struct net_device *dev, u8 tc);
 
 #ifdef CONFIG_RFS_ACCEL
 void mlx4_en_cleanup_filters(struct mlx4_en_priv *priv);