[PATCH net-next 0/4] r8169: improve HW csum and TSO handling

[PATCH net-next 0/4] r8169: improve HW csum and TSO handling

[Heiner Kallweit]

This series:
- delegates more tasks from the driver to the core
- enables HW csum and TSO per default
- copies quirks for buggy chip versions from vendor driver

Heiner Kallweit (4):
  r8169: set GSO size and segment limits
  r8169: implement callback ndo_features_check
  r8169: remove r8169_csum_workaround
  r8169: enable HW csum and TSO

 drivers/net/ethernet/realtek/r8169_main.c | 128 +++++++++++-----------
 1 file changed, 61 insertions(+), 67 deletions(-)

-- 
2.22.0

[PATCH net-next 1/4] r8169: set GSO size and segment limits

[Heiner Kallweit]

Set GSO max size and max segment number as in the vendor driver.

Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
---
 drivers/net/ethernet/realtek/r8169_main.c | 13 ++++++++++++-
 1 file changed, 12 insertions(+), 1 deletion(-)

diff --git a/drivers/net/ethernet/realtek/r8169_main.c b/drivers/net/ethernet/realtek/r8169_main.c
index 5c337234b..52a9e2d2f 100644
--- a/drivers/net/ethernet/realtek/r8169_main.c
+++ b/drivers/net/ethernet/realtek/r8169_main.c
@@ -569,6 +569,11 @@ enum rtl_rx_desc_bit {
 
 #define RsvdMask	0x3fffc000
 
+#define RTL_GSO_MAX_SIZE_V1	32000
+#define RTL_GSO_MAX_SEGS_V1	24
+#define RTL_GSO_MAX_SIZE_V2	64000
+#define RTL_GSO_MAX_SEGS_V2	64
+
 struct TxDesc {
 	__le32 opts1;
 	__le32 opts2;
@@ -6919,8 +6924,14 @@ static int rtl_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
 		/* Disallow toggling */
 		dev->hw_features &= ~NETIF_F_HW_VLAN_CTAG_RX;
 
-	if (rtl_chip_supports_csum_v2(tp))
+	if (rtl_chip_supports_csum_v2(tp)) {
 		dev->hw_features |= NETIF_F_IPV6_CSUM | NETIF_F_TSO6;
+		dev->gso_max_size = RTL_GSO_MAX_SIZE_V2;
+		dev->gso_max_segs = RTL_GSO_MAX_SEGS_V2;
+	} else {
+		dev->gso_max_size = RTL_GSO_MAX_SIZE_V1;
+		dev->gso_max_segs = RTL_GSO_MAX_SEGS_V1;
+	}
 
 	dev->hw_features |= NETIF_F_RXALL;
 	dev->hw_features |= NETIF_F_RXFCS;
-- 
2.22.0

[PATCH net-next 2/4] r8169: implement callback ndo_features_check

[Heiner Kallweit]

Implement callback ndo_features_check and move all feature checks there.
This will allow us to get rid of r8169_csum_workaround() completely in
a subsequent step. Like in the vendor driver disable HW csum for short
packets on RTL8168b.

Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
---
 drivers/net/ethernet/realtek/r8169_main.c | 60 ++++++++++++++---------
 1 file changed, 36 insertions(+), 24 deletions(-)

diff --git a/drivers/net/ethernet/realtek/r8169_main.c b/drivers/net/ethernet/realtek/r8169_main.c
index 52a9e2d2f..7e1b68b19 100644
--- a/drivers/net/ethernet/realtek/r8169_main.c
+++ b/drivers/net/ethernet/realtek/r8169_main.c
@@ -539,11 +539,11 @@ enum rtl_tx_desc_bit_1 {
 	TD1_GTSENV4	= (1 << 26),		/* Giant Send for IPv4 */
 	TD1_GTSENV6	= (1 << 25),		/* Giant Send for IPv6 */
 #define GTTCPHO_SHIFT			18
-#define GTTCPHO_MAX			0x7fU
+#define GTTCPHO_MAX			0x7f
 
 	/* Second doubleword. */
 #define TCPHO_SHIFT			18
-#define TCPHO_MAX			0x3ffU
+#define TCPHO_MAX			0x3ff
 #define TD1_MSS_SHIFT			18	/* MSS position (11 bits) */
 	TD1_IPv6_CS	= (1 << 28),		/* Calculate IPv6 checksum */
 	TD1_IPv4_CS	= (1 << 29),		/* Calculate IPv4 checksum */
@@ -5534,11 +5534,6 @@ static void r8169_csum_workaround(struct rtl8169_private *tp,
 		} while (segs);
 
 		dev_consume_skb_any(skb);
-	} else if (skb->ip_summed == CHECKSUM_PARTIAL) {
-		if (skb_checksum_help(skb) < 0)
-			goto drop;
-
-		rtl8169_start_xmit(skb, tp->dev);
 	} else {
 drop:
 		tp->dev->stats.tx_dropped++;
@@ -5595,13 +5590,6 @@ static bool rtl8169_tso_csum_v2(struct rtl8169_private *tp,
 	u32 mss = skb_shinfo(skb)->gso_size;
 
 	if (mss) {
-		if (transport_offset > GTTCPHO_MAX) {
-			netif_warn(tp, tx_err, tp->dev,
-				   "Invalid transport offset 0x%x for TSO\n",
-				   transport_offset);
-			return false;
-		}
-
 		switch (vlan_get_protocol(skb)) {
 		case htons(ETH_P_IP):
 			opts[0] |= TD1_GTSENV4;
@@ -5624,16 +5612,6 @@ static bool rtl8169_tso_csum_v2(struct rtl8169_private *tp,
 	} else if (skb->ip_summed == CHECKSUM_PARTIAL) {
 		u8 ip_protocol;
 
-		if (unlikely(rtl_test_hw_pad_bug(tp, skb)))
-			return !(skb_checksum_help(skb) || eth_skb_pad(skb));
-
-		if (transport_offset > TCPHO_MAX) {
-			netif_warn(tp, tx_err, tp->dev,
-				   "Invalid transport offset 0x%x\n",
-				   transport_offset);
-			return false;
-		}
-
 		switch (vlan_get_protocol(skb)) {
 		case htons(ETH_P_IP):
 			opts[1] |= TD1_IPv4_CS;
@@ -5790,6 +5768,39 @@ static netdev_tx_t rtl8169_start_xmit(struct sk_buff *skb,
 	return NETDEV_TX_BUSY;
 }
 
+static netdev_features_t rtl8169_features_check(struct sk_buff *skb,
+						struct net_device *dev,
+						netdev_features_t features)
+{
+	int transport_offset = skb_transport_offset(skb);
+	struct rtl8169_private *tp = netdev_priv(dev);
+
+	if (skb_is_gso(skb)) {
+		if (transport_offset > GTTCPHO_MAX &&
+		    rtl_chip_supports_csum_v2(tp))
+			features &= ~NETIF_F_ALL_TSO;
+	} else if (skb->ip_summed == CHECKSUM_PARTIAL) {
+		if (skb->len < ETH_ZLEN) {
+			switch (tp->mac_version) {
+			case RTL_GIGA_MAC_VER_11:
+			case RTL_GIGA_MAC_VER_12:
+			case RTL_GIGA_MAC_VER_17:
+			case RTL_GIGA_MAC_VER_34:
+				features &= ~NETIF_F_CSUM_MASK;
+				break;
+			default:
+				break;
+			}
+		}
+
+		if (transport_offset > TCPHO_MAX &&
+		    rtl_chip_supports_csum_v2(tp))
+			features &= ~NETIF_F_CSUM_MASK;
+	}
+
+	return vlan_features_check(skb, features);
+}
+
 static void rtl8169_pcierr_interrupt(struct net_device *dev)
 {
 	struct rtl8169_private *tp = netdev_priv(dev);
@@ -6546,6 +6557,7 @@ static const struct net_device_ops rtl_netdev_ops = {
 	.ndo_stop		= rtl8169_close,
 	.ndo_get_stats64	= rtl8169_get_stats64,
 	.ndo_start_xmit		= rtl8169_start_xmit,
+	.ndo_features_check	= rtl8169_features_check,
 	.ndo_tx_timeout		= rtl8169_tx_timeout,
 	.ndo_validate_addr	= eth_validate_addr,
 	.ndo_change_mtu		= rtl8169_change_mtu,
-- 
2.22.0

[PATCH net-next 3/4] r8169: remove r8169_csum_workaround

[Heiner Kallweit]

The loop in r8169_csum_workaround is called only if in
msdn_giant_send_check a copy of the skb header needs to be made and
we don't have enough memory. Let's simply drop the packet in that case
so that we can remove r8169_csum_workaround.

Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
---
 drivers/net/ethernet/realtek/r8169_main.c | 39 ++---------------------
 1 file changed, 2 insertions(+), 37 deletions(-)

diff --git a/drivers/net/ethernet/realtek/r8169_main.c b/drivers/net/ethernet/realtek/r8169_main.c
index 7e1b68b19..f77159540 100644
--- a/drivers/net/ethernet/realtek/r8169_main.c
+++ b/drivers/net/ethernet/realtek/r8169_main.c
@@ -5508,39 +5508,6 @@ static bool rtl_test_hw_pad_bug(struct rtl8169_private *tp, struct sk_buff *skb)
 	return skb->len < ETH_ZLEN && tp->mac_version == RTL_GIGA_MAC_VER_34;
 }
 
-static netdev_tx_t rtl8169_start_xmit(struct sk_buff *skb,
-				      struct net_device *dev);
-/* r8169_csum_workaround()
- * The hw limites the value the transport offset. When the offset is out of the
- * range, calculate the checksum by sw.
- */
-static void r8169_csum_workaround(struct rtl8169_private *tp,
-				  struct sk_buff *skb)
-{
-	if (skb_is_gso(skb)) {
-		netdev_features_t features = tp->dev->features;
-		struct sk_buff *segs, *nskb;
-
-		features &= ~(NETIF_F_SG | NETIF_F_IPV6_CSUM | NETIF_F_TSO6);
-		segs = skb_gso_segment(skb, features);
-		if (IS_ERR(segs) || !segs)
-			goto drop;
-
-		do {
-			nskb = segs;
-			segs = segs->next;
-			nskb->next = NULL;
-			rtl8169_start_xmit(nskb, tp->dev);
-		} while (segs);
-
-		dev_consume_skb_any(skb);
-	} else {
-drop:
-		tp->dev->stats.tx_dropped++;
-		dev_kfree_skb_any(skb);
-	}
-}
-
 /* msdn_giant_send_check()
  * According to the document of microsoft, the TCP Pseudo Header excludes the
  * packet length for IPv6 TCP large packets.
@@ -5688,10 +5655,8 @@ static netdev_tx_t rtl8169_start_xmit(struct sk_buff *skb,
 	opts[0] = DescOwn;
 
 	if (rtl_chip_supports_csum_v2(tp)) {
-		if (!rtl8169_tso_csum_v2(tp, skb, opts)) {
-			r8169_csum_workaround(tp, skb);
-			return NETDEV_TX_OK;
-		}
+		if (!rtl8169_tso_csum_v2(tp, skb, opts))
+			goto err_dma_0;
 	} else {
 		rtl8169_tso_csum_v1(skb, opts);
 	}
-- 
2.22.0

[PATCH net-next 4/4] r8169: enable HW csum and TSO

[Heiner Kallweit]

Enable HW csum and TSO per default except on known buggy chip versions.
Realtek confirmed that RTL8168evl has a HW issue with TSO.

Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
---
 drivers/net/ethernet/realtek/r8169_main.c | 16 +++++++++++-----
 1 file changed, 11 insertions(+), 5 deletions(-)

diff --git a/drivers/net/ethernet/realtek/r8169_main.c b/drivers/net/ethernet/realtek/r8169_main.c
index f77159540..61a23aee0 100644
--- a/drivers/net/ethernet/realtek/r8169_main.c
+++ b/drivers/net/ethernet/realtek/r8169_main.c
@@ -6879,11 +6879,9 @@ static int rtl_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
 
 	netif_napi_add(dev, &tp->napi, rtl8169_poll, NAPI_POLL_WEIGHT);
 
-	/* don't enable SG, IP_CSUM and TSO by default - it might not work
-	 * properly for all devices */
-	dev->features |= NETIF_F_RXCSUM |
-		NETIF_F_HW_VLAN_CTAG_TX | NETIF_F_HW_VLAN_CTAG_RX;
-
+	dev->features |= NETIF_F_SG | NETIF_F_IP_CSUM | NETIF_F_TSO |
+		NETIF_F_RXCSUM | NETIF_F_HW_VLAN_CTAG_TX |
+		NETIF_F_HW_VLAN_CTAG_RX;
 	dev->hw_features = NETIF_F_SG | NETIF_F_IP_CSUM | NETIF_F_TSO |
 		NETIF_F_RXCSUM | NETIF_F_HW_VLAN_CTAG_TX |
 		NETIF_F_HW_VLAN_CTAG_RX;
@@ -6903,6 +6901,7 @@ static int rtl_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
 
 	if (rtl_chip_supports_csum_v2(tp)) {
 		dev->hw_features |= NETIF_F_IPV6_CSUM | NETIF_F_TSO6;
+		dev->features |= NETIF_F_IPV6_CSUM | NETIF_F_TSO6;
 		dev->gso_max_size = RTL_GSO_MAX_SIZE_V2;
 		dev->gso_max_segs = RTL_GSO_MAX_SEGS_V2;
 	} else {
@@ -6910,6 +6909,13 @@ static int rtl_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
 		dev->gso_max_segs = RTL_GSO_MAX_SEGS_V1;
 	}
 
+	/* RTL8168e-vl has a HW issue with TSO */
+	if (tp->mac_version == RTL_GIGA_MAC_VER_34) {
+		dev->vlan_features &= ~NETIF_F_ALL_TSO;
+		dev->hw_features &= ~NETIF_F_ALL_TSO;
+		dev->features &= ~NETIF_F_ALL_TSO;
+	}
+
 	dev->hw_features |= NETIF_F_RXALL;
 	dev->hw_features |= NETIF_F_RXFCS;
 
-- 
2.22.0

Re: [PATCH net-next 0/4] r8169: improve HW csum and TSO handling

[David Miller]

From: Heiner Kallweit <hkallweit1@gmail.com>
Date: Fri, 26 Jul 2019 21:47:30 +0200

> This series:
> - delegates more tasks from the driver to the core
> - enables HW csum and TSO per default
> - copies quirks for buggy chip versions from vendor driver

Series applied, thanks Heiner.