[PATCH 1/5] net: ethernet: mediatek: add REVMII and fix RMII modes supported by GMAC

[PATCH 1/5] net: ethernet: mediatek: add REVMII and fix RMII modes supported by GMAC

[Sean Wang]

The patch adds the supplement for the setup of reverse MII (REVMII)
on GMAC, fixes up incorrect setup of reduced mii (RMII) on GMAC and
rearranges the error handling path for invalid phy argument.

Signed-off-by: Sean Wang <sean.wang@mediatek.com>
---
 drivers/net/ethernet/mediatek/mtk_eth_soc.c |   15 ++++++++++++---
 1 file changed, 12 insertions(+), 3 deletions(-)

diff --git a/drivers/net/ethernet/mediatek/mtk_eth_soc.c b/drivers/net/ethernet/mediatek/mtk_eth_soc.c
index 88b04dd..f19b8b9 100644
--- a/drivers/net/ethernet/mediatek/mtk_eth_soc.c
+++ b/drivers/net/ethernet/mediatek/mtk_eth_soc.c
@@ -245,12 +245,16 @@ static int mtk_phy_connect(struct mtk_mac *mac)
 	case PHY_INTERFACE_MODE_MII:
 		ge_mode = 1;
 		break;
-	case PHY_INTERFACE_MODE_RMII:
+	case PHY_INTERFACE_MODE_REVMII:
 		ge_mode = 2;
 		break;
+	case PHY_INTERFACE_MODE_RMII:
+		if (!mac->id)
+			goto err_phy;
+		ge_mode = 3;
+		break;
 	default:
-		dev_err(eth->dev, "invalid phy_mode\n");
-		return -1;
+		goto err_phy;
 	}
 
 	/* put the gmac into the right mode */
@@ -272,6 +276,11 @@ static int mtk_phy_connect(struct mtk_mac *mac)
 	of_node_put(np);
 
 	return 0;
+
+err_phy:
+	of_node_put(np);
+	dev_err(eth->dev, "invalid phy_mode\n");
+	return -EINVAL;
 }
 
 static int mtk_mdio_init(struct mtk_eth *eth)
-- 
1.7.9.5

[PATCH 1/5] net: ethernet: mediatek: add REVMII and fix RMII modes supported by GMAC

[Sean Wang]

The patch adds the supplement for the setup of reverse MII (REVMII)
on GMAC, fixes up incorrect setup of reduced mii (RMII) on GMAC and
rearranges the error handling path for invalid phy argument.

Signed-off-by: Sean Wang <sean.wang@mediatek.com>
---
 drivers/net/ethernet/mediatek/mtk_eth_soc.c |   15 ++++++++++++---
 1 file changed, 12 insertions(+), 3 deletions(-)

diff --git a/drivers/net/ethernet/mediatek/mtk_eth_soc.c b/drivers/net/ethernet/mediatek/mtk_eth_soc.c
index 88b04dd..f19b8b9 100644
--- a/drivers/net/ethernet/mediatek/mtk_eth_soc.c
+++ b/drivers/net/ethernet/mediatek/mtk_eth_soc.c
@@ -245,12 +245,16 @@ static int mtk_phy_connect(struct mtk_mac *mac)
 	case PHY_INTERFACE_MODE_MII:
 		ge_mode = 1;
 		break;
-	case PHY_INTERFACE_MODE_RMII:
+	case PHY_INTERFACE_MODE_REVMII:
 		ge_mode = 2;
 		break;
+	case PHY_INTERFACE_MODE_RMII:
+		if (!mac->id)
+			goto err_phy;
+		ge_mode = 3;
+		break;
 	default:
-		dev_err(eth->dev, "invalid phy_mode\n");
-		return -1;
+		goto err_phy;
 	}
 
 	/* put the gmac into the right mode */
@@ -272,6 +276,11 @@ static int mtk_phy_connect(struct mtk_mac *mac)
 	of_node_put(np);
 
 	return 0;
+
+err_phy:
+	of_node_put(np);
+	dev_err(eth->dev, "invalid phy_mode\n");
+	return -EINVAL;
 }
 
 static int mtk_mdio_init(struct mtk_eth *eth)
-- 
1.7.9.5

[PATCH 2/5] net: ethernet: mediatek: fixed flow control settings on GMAC0 is not being enabled properly

[Sean Wang]

Commit 08ef55c6f257acf3bdc6940813f80e8f0f5d90ec
("net-next: mediatek: fix gigabit and flow control advertisement")
had supported proper flow control settings for GMAC1. But for GMAC0,

1.GMAC0 shares the common logic with GMAC1 inside mtk_phy_link_adjust()
to adapt various settings for the target phy.

2.GMAC1 uses fixed-phy to connect to a builtin gigabit switch with
fixed link speed as commit 0c72c50f6f93b0c3daa9ea35d89ab3a933c7b5a0
("net-next: mediatek: add fixed-phy support") describes.

3.However, fixed-phy doesn't enable SUPPORTED_Pause & SUPPORTED_Asym_Pause
supported flag on default that would cause mtk_phy_link_adjust() not to
enable flow control setting on GMAC0 properly and cause packet dropped
when high traffic.

Due to these reasons, the patch adds SUPPORTED_Pause & SUPPORTED_Asym_Pause
supported flags on fixed-phy used by driver to have proper handling on
the both GMAC with the shared common logic.

Signed-off-by: Sean Wang <sean.wang@mediatek.com>
---
 drivers/net/ethernet/mediatek/mtk_eth_soc.c |    5 +++++
 1 file changed, 5 insertions(+)

diff --git a/drivers/net/ethernet/mediatek/mtk_eth_soc.c b/drivers/net/ethernet/mediatek/mtk_eth_soc.c
index f19b8b9..eb59a73 100644
--- a/drivers/net/ethernet/mediatek/mtk_eth_soc.c
+++ b/drivers/net/ethernet/mediatek/mtk_eth_soc.c
@@ -267,6 +267,11 @@ static int mtk_phy_connect(struct mtk_mac *mac)
 	mac->phy_dev->autoneg = AUTONEG_ENABLE;
 	mac->phy_dev->speed = 0;
 	mac->phy_dev->duplex = 0;
+
+	if (of_phy_is_fixed_link(mac->of_node))
+		mac->phy_dev->supported |=
+		SUPPORTED_Pause | SUPPORTED_Asym_Pause;
+
 	mac->phy_dev->supported &= PHY_GBIT_FEATURES | SUPPORTED_Pause |
 				   SUPPORTED_Asym_Pause;
 	mac->phy_dev->advertising = mac->phy_dev->supported |
-- 
1.7.9.5

[PATCH 2/5] net: ethernet: mediatek: fixed flow control settings on GMAC0 is not being enabled properly

[Sean Wang]

Commit 08ef55c6f257acf3bdc6940813f80e8f0f5d90ec
("net-next: mediatek: fix gigabit and flow control advertisement")
had supported proper flow control settings for GMAC1. But for GMAC0,

1.GMAC0 shares the common logic with GMAC1 inside mtk_phy_link_adjust()
to adapt various settings for the target phy.

2.GMAC1 uses fixed-phy to connect to a builtin gigabit switch with
fixed link speed as commit 0c72c50f6f93b0c3daa9ea35d89ab3a933c7b5a0
("net-next: mediatek: add fixed-phy support") describes.

3.However, fixed-phy doesn't enable SUPPORTED_Pause & SUPPORTED_Asym_Pause
supported flag on default that would cause mtk_phy_link_adjust() not to
enable flow control setting on GMAC0 properly and cause packet dropped
when high traffic.

Due to these reasons, the patch adds SUPPORTED_Pause & SUPPORTED_Asym_Pause
supported flags on fixed-phy used by driver to have proper handling on
the both GMAC with the shared common logic.

Signed-off-by: Sean Wang <sean.wang@mediatek.com>
---
 drivers/net/ethernet/mediatek/mtk_eth_soc.c |    5 +++++
 1 file changed, 5 insertions(+)

diff --git a/drivers/net/ethernet/mediatek/mtk_eth_soc.c b/drivers/net/ethernet/mediatek/mtk_eth_soc.c
index f19b8b9..eb59a73 100644
--- a/drivers/net/ethernet/mediatek/mtk_eth_soc.c
+++ b/drivers/net/ethernet/mediatek/mtk_eth_soc.c
@@ -267,6 +267,11 @@ static int mtk_phy_connect(struct mtk_mac *mac)
 	mac->phy_dev->autoneg = AUTONEG_ENABLE;
 	mac->phy_dev->speed = 0;
 	mac->phy_dev->duplex = 0;
+
+	if (of_phy_is_fixed_link(mac->of_node))
+		mac->phy_dev->supported |=
+		SUPPORTED_Pause | SUPPORTED_Asym_Pause;
+
 	mac->phy_dev->supported &= PHY_GBIT_FEATURES | SUPPORTED_Pause |
 				   SUPPORTED_Asym_Pause;
 	mac->phy_dev->advertising = mac->phy_dev->supported |
-- 
1.7.9.5

[PATCH 3/5] net: ethernet: mediatek: fixed runtime warning raised by inconsistent struct device pointers passed to DMA API

[Sean Wang]

Runtime warning occurs if DMA-API debug feature is enabled that would be
raised by pointers passed to DMA API as arguments to inconsistent struct
device objects, so that the patch makes them usage aligned between dma
operations such as dma_map_*() and dma_unmap_*() to eliminate the warning.

Signed-off-by: Sean Wang <sean.wang@mediatek.com>
---
 drivers/net/ethernet/mediatek/mtk_eth_soc.c |   26 +++++++++++++-------------
 1 file changed, 13 insertions(+), 13 deletions(-)

diff --git a/drivers/net/ethernet/mediatek/mtk_eth_soc.c b/drivers/net/ethernet/mediatek/mtk_eth_soc.c
index eb59a73..b782330 100644
--- a/drivers/net/ethernet/mediatek/mtk_eth_soc.c
+++ b/drivers/net/ethernet/mediatek/mtk_eth_soc.c
@@ -557,15 +557,15 @@ static inline struct mtk_tx_buf *mtk_desc_to_tx_buf(struct mtk_tx_ring *ring,
 	return &ring->buf[idx];
 }
 
-static void mtk_tx_unmap(struct device *dev, struct mtk_tx_buf *tx_buf)
+static void mtk_tx_unmap(struct mtk_eth *eth, struct mtk_tx_buf *tx_buf)
 {
 	if (tx_buf->flags & MTK_TX_FLAGS_SINGLE0) {
-		dma_unmap_single(dev,
+		dma_unmap_single(eth->dev,
 				 dma_unmap_addr(tx_buf, dma_addr0),
 				 dma_unmap_len(tx_buf, dma_len0),
 				 DMA_TO_DEVICE);
 	} else if (tx_buf->flags & MTK_TX_FLAGS_PAGE0) {
-		dma_unmap_page(dev,
+		dma_unmap_page(eth->dev,
 			       dma_unmap_addr(tx_buf, dma_addr0),
 			       dma_unmap_len(tx_buf, dma_len0),
 			       DMA_TO_DEVICE);
@@ -610,9 +610,9 @@ static int mtk_tx_map(struct sk_buff *skb, struct net_device *dev,
 	if (skb_vlan_tag_present(skb))
 		txd4 |= TX_DMA_INS_VLAN | skb_vlan_tag_get(skb);
 
-	mapped_addr = dma_map_single(&dev->dev, skb->data,
+	mapped_addr = dma_map_single(eth->dev, skb->data,
 				     skb_headlen(skb), DMA_TO_DEVICE);
-	if (unlikely(dma_mapping_error(&dev->dev, mapped_addr)))
+	if (unlikely(dma_mapping_error(eth->dev, mapped_addr)))
 		return -ENOMEM;
 
 	WRITE_ONCE(itxd->txd1, mapped_addr);
@@ -638,10 +638,10 @@ static int mtk_tx_map(struct sk_buff *skb, struct net_device *dev,
 
 			n_desc++;
 			frag_map_size = min(frag_size, MTK_TX_DMA_BUF_LEN);
-			mapped_addr = skb_frag_dma_map(&dev->dev, frag, offset,
+			mapped_addr = skb_frag_dma_map(eth->dev, frag, offset,
 						       frag_map_size,
 						       DMA_TO_DEVICE);
-			if (unlikely(dma_mapping_error(&dev->dev, mapped_addr)))
+			if (unlikely(dma_mapping_error(eth->dev, mapped_addr)))
 				goto err_dma;
 
 			if (i == nr_frags - 1 &&
@@ -694,7 +694,7 @@ err_dma:
 		tx_buf = mtk_desc_to_tx_buf(ring, itxd);
 
 		/* unmap dma */
-		mtk_tx_unmap(&dev->dev, tx_buf);
+		mtk_tx_unmap(eth, tx_buf);
 
 		itxd->txd3 = TX_DMA_LS0 | TX_DMA_OWNER_CPU;
 		itxd = mtk_qdma_phys_to_virt(ring, itxd->txd2);
@@ -850,11 +850,11 @@ static int mtk_poll_rx(struct napi_struct *napi, int budget,
 			netdev->stats.rx_dropped++;
 			goto release_desc;
 		}
-		dma_addr = dma_map_single(&eth->netdev[mac]->dev,
+		dma_addr = dma_map_single(eth->dev,
 					  new_data + NET_SKB_PAD,
 					  ring->buf_size,
 					  DMA_FROM_DEVICE);
-		if (unlikely(dma_mapping_error(&netdev->dev, dma_addr))) {
+		if (unlikely(dma_mapping_error(eth->dev, dma_addr))) {
 			skb_free_frag(new_data);
 			netdev->stats.rx_dropped++;
 			goto release_desc;
@@ -869,7 +869,7 @@ static int mtk_poll_rx(struct napi_struct *napi, int budget,
 		}
 		skb_reserve(skb, NET_SKB_PAD + NET_IP_ALIGN);
 
-		dma_unmap_single(&netdev->dev, trxd.rxd1,
+		dma_unmap_single(eth->dev, trxd.rxd1,
 				 ring->buf_size, DMA_FROM_DEVICE);
 		pktlen = RX_DMA_GET_PLEN0(trxd.rxd2);
 		skb->dev = netdev;
@@ -951,7 +951,7 @@ static int mtk_poll_tx(struct mtk_eth *eth, int budget)
 			done[mac]++;
 			budget--;
 		}
-		mtk_tx_unmap(eth->dev, tx_buf);
+		mtk_tx_unmap(eth, tx_buf);
 
 		ring->last_free = desc;
 		atomic_inc(&ring->free_count);
@@ -1106,7 +1106,7 @@ static void mtk_tx_clean(struct mtk_eth *eth)
 
 	if (ring->buf) {
 		for (i = 0; i < MTK_DMA_SIZE; i++)
-			mtk_tx_unmap(eth->dev, &ring->buf[i]);
+			mtk_tx_unmap(eth, &ring->buf[i]);
 		kfree(ring->buf);
 		ring->buf = NULL;
 	}
-- 
1.7.9.5

[PATCH 3/5] net: ethernet: mediatek: fixed runtime warning raised by inconsistent struct device pointers passed to DMA API

[Sean Wang]

Runtime warning occurs if DMA-API debug feature is enabled that would be
raised by pointers passed to DMA API as arguments to inconsistent struct
device objects, so that the patch makes them usage aligned between dma
operations such as dma_map_*() and dma_unmap_*() to eliminate the warning.

Signed-off-by: Sean Wang <sean.wang@mediatek.com>
---
 drivers/net/ethernet/mediatek/mtk_eth_soc.c |   26 +++++++++++++-------------
 1 file changed, 13 insertions(+), 13 deletions(-)

diff --git a/drivers/net/ethernet/mediatek/mtk_eth_soc.c b/drivers/net/ethernet/mediatek/mtk_eth_soc.c
index eb59a73..b782330 100644
--- a/drivers/net/ethernet/mediatek/mtk_eth_soc.c
+++ b/drivers/net/ethernet/mediatek/mtk_eth_soc.c
@@ -557,15 +557,15 @@ static inline struct mtk_tx_buf *mtk_desc_to_tx_buf(struct mtk_tx_ring *ring,
 	return &ring->buf[idx];
 }
 
-static void mtk_tx_unmap(struct device *dev, struct mtk_tx_buf *tx_buf)
+static void mtk_tx_unmap(struct mtk_eth *eth, struct mtk_tx_buf *tx_buf)
 {
 	if (tx_buf->flags & MTK_TX_FLAGS_SINGLE0) {
-		dma_unmap_single(dev,
+		dma_unmap_single(eth->dev,
 				 dma_unmap_addr(tx_buf, dma_addr0),
 				 dma_unmap_len(tx_buf, dma_len0),
 				 DMA_TO_DEVICE);
 	} else if (tx_buf->flags & MTK_TX_FLAGS_PAGE0) {
-		dma_unmap_page(dev,
+		dma_unmap_page(eth->dev,
 			       dma_unmap_addr(tx_buf, dma_addr0),
 			       dma_unmap_len(tx_buf, dma_len0),
 			       DMA_TO_DEVICE);
@@ -610,9 +610,9 @@ static int mtk_tx_map(struct sk_buff *skb, struct net_device *dev,
 	if (skb_vlan_tag_present(skb))
 		txd4 |= TX_DMA_INS_VLAN | skb_vlan_tag_get(skb);
 
-	mapped_addr = dma_map_single(&dev->dev, skb->data,
+	mapped_addr = dma_map_single(eth->dev, skb->data,
 				     skb_headlen(skb), DMA_TO_DEVICE);
-	if (unlikely(dma_mapping_error(&dev->dev, mapped_addr)))
+	if (unlikely(dma_mapping_error(eth->dev, mapped_addr)))
 		return -ENOMEM;
 
 	WRITE_ONCE(itxd->txd1, mapped_addr);
@@ -638,10 +638,10 @@ static int mtk_tx_map(struct sk_buff *skb, struct net_device *dev,
 
 			n_desc++;
 			frag_map_size = min(frag_size, MTK_TX_DMA_BUF_LEN);
-			mapped_addr = skb_frag_dma_map(&dev->dev, frag, offset,
+			mapped_addr = skb_frag_dma_map(eth->dev, frag, offset,
 						       frag_map_size,
 						       DMA_TO_DEVICE);
-			if (unlikely(dma_mapping_error(&dev->dev, mapped_addr)))
+			if (unlikely(dma_mapping_error(eth->dev, mapped_addr)))
 				goto err_dma;
 
 			if (i == nr_frags - 1 &&
@@ -694,7 +694,7 @@ err_dma:
 		tx_buf = mtk_desc_to_tx_buf(ring, itxd);
 
 		/* unmap dma */
-		mtk_tx_unmap(&dev->dev, tx_buf);
+		mtk_tx_unmap(eth, tx_buf);
 
 		itxd->txd3 = TX_DMA_LS0 | TX_DMA_OWNER_CPU;
 		itxd = mtk_qdma_phys_to_virt(ring, itxd->txd2);
@@ -850,11 +850,11 @@ static int mtk_poll_rx(struct napi_struct *napi, int budget,
 			netdev->stats.rx_dropped++;
 			goto release_desc;
 		}
-		dma_addr = dma_map_single(&eth->netdev[mac]->dev,
+		dma_addr = dma_map_single(eth->dev,
 					  new_data + NET_SKB_PAD,
 					  ring->buf_size,
 					  DMA_FROM_DEVICE);
-		if (unlikely(dma_mapping_error(&netdev->dev, dma_addr))) {
+		if (unlikely(dma_mapping_error(eth->dev, dma_addr))) {
 			skb_free_frag(new_data);
 			netdev->stats.rx_dropped++;
 			goto release_desc;
@@ -869,7 +869,7 @@ static int mtk_poll_rx(struct napi_struct *napi, int budget,
 		}
 		skb_reserve(skb, NET_SKB_PAD + NET_IP_ALIGN);
 
-		dma_unmap_single(&netdev->dev, trxd.rxd1,
+		dma_unmap_single(eth->dev, trxd.rxd1,
 				 ring->buf_size, DMA_FROM_DEVICE);
 		pktlen = RX_DMA_GET_PLEN0(trxd.rxd2);
 		skb->dev = netdev;
@@ -951,7 +951,7 @@ static int mtk_poll_tx(struct mtk_eth *eth, int budget)
 			done[mac]++;
 			budget--;
 		}
-		mtk_tx_unmap(eth->dev, tx_buf);
+		mtk_tx_unmap(eth, tx_buf);
 
 		ring->last_free = desc;
 		atomic_inc(&ring->free_count);
@@ -1106,7 +1106,7 @@ static void mtk_tx_clean(struct mtk_eth *eth)
 
 	if (ring->buf) {
 		for (i = 0; i < MTK_DMA_SIZE; i++)
-			mtk_tx_unmap(eth->dev, &ring->buf[i]);
+			mtk_tx_unmap(eth, &ring->buf[i]);
 		kfree(ring->buf);
 		ring->buf = NULL;
 	}
-- 
1.7.9.5

[PATCH 4/5] net: ethernet: mediatek: enable rx path by reducing memory barrier

[Sean Wang]

The patch makes moving wmb() to outside the loop that could help
rx path handling more faster although that rx descriptors aren't
freed for DMA to use as soon as possible, but based on my experiment
and the result show it still can reach about 943Mbpis tested on the
setup with one port using gigaphy and 256 rx descriptors for
DMA to move.

Signed-off-by: Sean Wang <sean.wang@mediatek.com>
---
 drivers/net/ethernet/mediatek/mtk_eth_soc.c |   11 ++++++-----
 1 file changed, 6 insertions(+), 5 deletions(-)

diff --git a/drivers/net/ethernet/mediatek/mtk_eth_soc.c b/drivers/net/ethernet/mediatek/mtk_eth_soc.c
index b782330..53e24c1 100644
--- a/drivers/net/ethernet/mediatek/mtk_eth_soc.c
+++ b/drivers/net/ethernet/mediatek/mtk_eth_soc.c
@@ -893,14 +893,15 @@ release_desc:
 		rxd->rxd2 = RX_DMA_PLEN0(ring->buf_size);
 
 		ring->calc_idx = idx;
-		/* make sure that all changes to the dma ring are flushed before
-		 * we continue
-		 */
-		wmb();
-		mtk_w32(eth, ring->calc_idx, MTK_QRX_CRX_IDX0);
 		done++;
 	}
 
+    /* make sure that all changes to the dma ring are flushed before
+     * we continue
+     */
+	wmb();
+	mtk_w32(eth, ring->calc_idx, MTK_QRX_CRX_IDX0);
+
 	if (done < budget)
 		mtk_w32(eth, MTK_RX_DONE_INT, MTK_QMTK_INT_STATUS);
 
-- 
1.7.9.5

[PATCH 4/5] net: ethernet: mediatek: enable rx path by reducing memory barrier

[Sean Wang]

The patch makes moving wmb() to outside the loop that could help
rx path handling more faster although that rx descriptors aren't
freed for DMA to use as soon as possible, but based on my experiment
and the result show it still can reach about 943Mbpis tested on the
setup with one port using gigaphy and 256 rx descriptors for
DMA to move.

Signed-off-by: Sean Wang <sean.wang@mediatek.com>
---
 drivers/net/ethernet/mediatek/mtk_eth_soc.c |   11 ++++++-----
 1 file changed, 6 insertions(+), 5 deletions(-)

diff --git a/drivers/net/ethernet/mediatek/mtk_eth_soc.c b/drivers/net/ethernet/mediatek/mtk_eth_soc.c
index b782330..53e24c1 100644
--- a/drivers/net/ethernet/mediatek/mtk_eth_soc.c
+++ b/drivers/net/ethernet/mediatek/mtk_eth_soc.c
@@ -893,14 +893,15 @@ release_desc:
 		rxd->rxd2 = RX_DMA_PLEN0(ring->buf_size);
 
 		ring->calc_idx = idx;
-		/* make sure that all changes to the dma ring are flushed before
-		 * we continue
-		 */
-		wmb();
-		mtk_w32(eth, ring->calc_idx, MTK_QRX_CRX_IDX0);
 		done++;
 	}
 
+    /* make sure that all changes to the dma ring are flushed before
+     * we continue
+     */
+	wmb();
+	mtk_w32(eth, ring->calc_idx, MTK_QRX_CRX_IDX0);
+
 	if (done < budget)
 		mtk_w32(eth, MTK_RX_DONE_INT, MTK_QMTK_INT_STATUS);
 
-- 
1.7.9.5

[PATCH 5/5] net: ethernet: mediatek: enable rx path by aggregrating more skbs

[Sean Wang]

The patch adds support for aggregrating more skbs feed into NAPI in
order to get benifits from generic receive offload (GRO) by peeking
at the RX ring status and moving more packet right before returning
from NAPI rx handler if NAPI budgets is still available.

Signed-off-by: Sean Wang <sean.wang@mediatek.com>
---
 drivers/net/ethernet/mediatek/mtk_eth_soc.c |   15 +++++++--------
 1 file changed, 7 insertions(+), 8 deletions(-)

diff --git a/drivers/net/ethernet/mediatek/mtk_eth_soc.c b/drivers/net/ethernet/mediatek/mtk_eth_soc.c
index 53e24c1..742b9ec 100644
--- a/drivers/net/ethernet/mediatek/mtk_eth_soc.c
+++ b/drivers/net/ethernet/mediatek/mtk_eth_soc.c
@@ -902,9 +902,6 @@ release_desc:
 	wmb();
 	mtk_w32(eth, ring->calc_idx, MTK_QRX_CRX_IDX0);
 
-	if (done < budget)
-		mtk_w32(eth, MTK_RX_DONE_INT, MTK_QMTK_INT_STATUS);
-
 	return done;
 }
 
@@ -1023,8 +1020,10 @@ static int mtk_napi_rx(struct napi_struct *napi, int budget)
 	struct mtk_eth *eth = container_of(napi, struct mtk_eth, rx_napi);
 	u32 status, mask;
 	int rx_done = 0;
+	int remain_budget = budget;
 
 	mtk_handle_status_irq(eth);
+poll_again:
 	mtk_w32(eth, MTK_RX_DONE_INT, MTK_QMTK_INT_STATUS);
 	rx_done = mtk_poll_rx(napi, budget, eth);
 
@@ -1035,14 +1034,14 @@ static int mtk_napi_rx(struct napi_struct *napi, int budget)
 			 "done rx %d, intr 0x%08x/0x%x\n",
 			 rx_done, status, mask);
 	}
-
-	if (rx_done == budget)
+	if (rx_done == remain_budget)
 		return budget;
 
 	status = mtk_r32(eth, MTK_QMTK_INT_STATUS);
-	if (status & MTK_RX_DONE_INT)
-		return budget;
-
+	if (status & MTK_RX_DONE_INT) {
+		remain_budget -= rx_done;
+		goto poll_again;
+	}
 	napi_complete(napi);
 	mtk_irq_enable(eth, MTK_RX_DONE_INT);
 
-- 
1.7.9.5

[PATCH 5/5] net: ethernet: mediatek: enable rx path by aggregrating more skbs

[Sean Wang]

The patch adds support for aggregrating more skbs feed into NAPI in
order to get benifits from generic receive offload (GRO) by peeking
at the RX ring status and moving more packet right before returning
from NAPI rx handler if NAPI budgets is still available.

Signed-off-by: Sean Wang <sean.wang@mediatek.com>
---
 drivers/net/ethernet/mediatek/mtk_eth_soc.c |   15 +++++++--------
 1 file changed, 7 insertions(+), 8 deletions(-)

diff --git a/drivers/net/ethernet/mediatek/mtk_eth_soc.c b/drivers/net/ethernet/mediatek/mtk_eth_soc.c
index 53e24c1..742b9ec 100644
--- a/drivers/net/ethernet/mediatek/mtk_eth_soc.c
+++ b/drivers/net/ethernet/mediatek/mtk_eth_soc.c
@@ -902,9 +902,6 @@ release_desc:
 	wmb();
 	mtk_w32(eth, ring->calc_idx, MTK_QRX_CRX_IDX0);
 
-	if (done < budget)
-		mtk_w32(eth, MTK_RX_DONE_INT, MTK_QMTK_INT_STATUS);
-
 	return done;
 }
 
@@ -1023,8 +1020,10 @@ static int mtk_napi_rx(struct napi_struct *napi, int budget)
 	struct mtk_eth *eth = container_of(napi, struct mtk_eth, rx_napi);
 	u32 status, mask;
 	int rx_done = 0;
+	int remain_budget = budget;
 
 	mtk_handle_status_irq(eth);
+poll_again:
 	mtk_w32(eth, MTK_RX_DONE_INT, MTK_QMTK_INT_STATUS);
 	rx_done = mtk_poll_rx(napi, budget, eth);
 
@@ -1035,14 +1034,14 @@ static int mtk_napi_rx(struct napi_struct *napi, int budget)
 			 "done rx %d, intr 0x%08x/0x%x\n",
 			 rx_done, status, mask);
 	}
-
-	if (rx_done == budget)
+	if (rx_done == remain_budget)
 		return budget;
 
 	status = mtk_r32(eth, MTK_QMTK_INT_STATUS);
-	if (status & MTK_RX_DONE_INT)
-		return budget;
-
+	if (status & MTK_RX_DONE_INT) {
+		remain_budget -= rx_done;
+		goto poll_again;
+	}
 	napi_complete(napi);
 	mtk_irq_enable(eth, MTK_RX_DONE_INT);
 
-- 
1.7.9.5

Re: [PATCH 4/5] net: ethernet: mediatek: enable rx path by reducing memory barrier

[Sergei Shtylyov]

Hello.

On 08/15/2016 06:03 PM, Sean Wang wrote:

> The patch makes moving wmb() to outside the loop that could help
> rx path handling more faster although that rx descriptors aren't

    RX.

> freed for DMA to use as soon as possible, but based on my experiment
> and the result show it still can reach about 943Mbpis tested on the

    Shows.

> setup with one port using gigaphy and 256 rx descriptors for

    RX.

> DMA to move.
>
> Signed-off-by: Sean Wang <sean.wang-NuS5LvNUpcJWk0Htik3J/w@public.gmane.org>
> ---
>  drivers/net/ethernet/mediatek/mtk_eth_soc.c |   11 ++++++-----
>  1 file changed, 6 insertions(+), 5 deletions(-)
>
> diff --git a/drivers/net/ethernet/mediatek/mtk_eth_soc.c b/drivers/net/ethernet/mediatek/mtk_eth_soc.c
> index b782330..53e24c1 100644
> --- a/drivers/net/ethernet/mediatek/mtk_eth_soc.c
> +++ b/drivers/net/ethernet/mediatek/mtk_eth_soc.c
> @@ -893,14 +893,15 @@ release_desc:
>  		rxd->rxd2 = RX_DMA_PLEN0(ring->buf_size);
>
>  		ring->calc_idx = idx;
> -		/* make sure that all changes to the dma ring are flushed before
> -		 * we continue
> -		 */
> -		wmb();
> -		mtk_w32(eth, ring->calc_idx, MTK_QRX_CRX_IDX0);
>  		done++;
>  	}
>
> +    /* make sure that all changes to the dma ring are flushed before
> +     * we continue
> +     */

    Please indent the comment with a tab, like the code after it.

> +	wmb();
> +	mtk_w32(eth, ring->calc_idx, MTK_QRX_CRX_IDX0);
> +
>  	if (done < budget)
>  		mtk_w32(eth, MTK_RX_DONE_INT, MTK_QMTK_INT_STATUS);
>

MBR, Sergei