[PATCH V2 net-next 0/5] net: hns3: Add support of hardware GRO to HNS3 Driver

[PATCH V2 net-next 0/5] net: hns3: Add support of hardware GRO to HNS3 Driver

[Salil Mehta]

This patch-set adds support of hardware assisted GRO feature to
HNS3 driver on Rev B(=0x21) platform. Current hardware only
supports TCP/IPv{4|6} flows.

Change Log:
V1->V2:
1. Remove redundant print reported by Leon Romanovsky.
   Link: https://lkml.org/lkml/2018/11/13/715

Peng Li (5):
  net: hns3: Enable HW GRO for Rev B(=0x21) HNS3 hardware
  net: hns3: Add handling of GRO Pkts not fully RX'ed in NAPI poll
  net: hns3: Add support for ethtool -K to enable/disable HW GRO
  net: hns3: Add skb chain when num of RX buf exceeds MAX_SKB_FRAGS
  net: hns3: Adds GRO params to SKB for the stack

 drivers/net/ethernet/hisilicon/hns3/hnae3.h   |   7 +
 .../net/ethernet/hisilicon/hns3/hns3_enet.c   | 289 ++++++++++++++----
 .../net/ethernet/hisilicon/hns3/hns3_enet.h   |  17 +-
 .../hisilicon/hns3/hns3pf/hclge_cmd.h         |   7 +
 .../hisilicon/hns3/hns3pf/hclge_main.c        |  39 +++
 .../hisilicon/hns3/hns3vf/hclgevf_cmd.h       |   8 +
 .../hisilicon/hns3/hns3vf/hclgevf_main.c      |  39 +++
 7 files changed, 339 insertions(+), 67 deletions(-)

-- 
2.17.1

[PATCH V2 net-next 1/5] net: hns3: Enable HW GRO for Rev B(=0x21) HNS3 hardware

[Salil Mehta]

From: Peng Li <lipeng321@huawei.com>

HNS3 hardware Revision B(=0x21) supports Hardware GRO feature. This
patch enables this feature in the HNS3 PF/VF driver.

Signed-off-by: Peng Li <lipeng321@huawei.com>
Signed-off-by: Salil Mehta <salil.mehta@huawei.com>
---
 drivers/net/ethernet/hisilicon/hns3/hnae3.h   |  4 +++
 .../net/ethernet/hisilicon/hns3/hns3_enet.c   |  4 ++-
 .../hisilicon/hns3/hns3pf/hclge_cmd.h         |  7 +++++
 .../hisilicon/hns3/hns3pf/hclge_main.c        | 30 ++++++++++++++++++
 .../hisilicon/hns3/hns3vf/hclgevf_cmd.h       |  8 +++++
 .../hisilicon/hns3/hns3vf/hclgevf_main.c      | 31 +++++++++++++++++++
 6 files changed, 83 insertions(+), 1 deletion(-)

diff --git a/drivers/net/ethernet/hisilicon/hns3/hnae3.h b/drivers/net/ethernet/hisilicon/hns3/hnae3.h
index f69d39f17bdd..21d934b7a2a3 100644
--- a/drivers/net/ethernet/hisilicon/hns3/hnae3.h
+++ b/drivers/net/ethernet/hisilicon/hns3/hnae3.h
@@ -52,6 +52,7 @@
 #define HNAE3_UNIC_CLIENT_INITED_B		0x4
 #define HNAE3_ROCE_CLIENT_INITED_B		0x5
 #define HNAE3_DEV_SUPPORT_FD_B			0x6
+#define HNAE3_DEV_SUPPORT_GRO_B			0x7
 
 #define HNAE3_DEV_SUPPORT_ROCE_DCB_BITS (BIT(HNAE3_DEV_SUPPORT_DCB_B) |\
 		BIT(HNAE3_DEV_SUPPORT_ROCE_B))
@@ -65,6 +66,9 @@
 #define hnae3_dev_fd_supported(hdev) \
 	hnae3_get_bit((hdev)->ae_dev->flag, HNAE3_DEV_SUPPORT_FD_B)
 
+#define hnae3_dev_gro_supported(hdev) \
+	hnae3_get_bit((hdev)->ae_dev->flag, HNAE3_DEV_SUPPORT_GRO_B)
+
 #define ring_ptr_move_fw(ring, p) \
 	((ring)->p = ((ring)->p + 1) % (ring)->desc_num)
 #define ring_ptr_move_bw(ring, p) \
diff --git a/drivers/net/ethernet/hisilicon/hns3/hns3_enet.c b/drivers/net/ethernet/hisilicon/hns3/hns3_enet.c
index 8d07ec668d5d..a510ddfd45a5 100644
--- a/drivers/net/ethernet/hisilicon/hns3/hns3_enet.c
+++ b/drivers/net/ethernet/hisilicon/hns3/hns3_enet.c
@@ -1714,8 +1714,10 @@ static void hns3_disable_sriov(struct pci_dev *pdev)
 static void hns3_get_dev_capability(struct pci_dev *pdev,
 				    struct hnae3_ae_dev *ae_dev)
 {
-	if (pdev->revision >= 0x21)
+	if (pdev->revision >= 0x21) {
 		hnae3_set_bit(ae_dev->flag, HNAE3_DEV_SUPPORT_FD_B, 1);
+		hnae3_set_bit(ae_dev->flag, HNAE3_DEV_SUPPORT_GRO_B, 1);
+	}
 }
 
 /* hns3_probe - Device initialization routine
diff --git a/drivers/net/ethernet/hisilicon/hns3/hns3pf/hclge_cmd.h b/drivers/net/ethernet/hisilicon/hns3/hns3pf/hclge_cmd.h
index 872cd4bdd70d..aef044d08b11 100644
--- a/drivers/net/ethernet/hisilicon/hns3/hns3pf/hclge_cmd.h
+++ b/drivers/net/ethernet/hisilicon/hns3/hns3pf/hclge_cmd.h
@@ -152,6 +152,7 @@ enum hclge_opcode_type {
 
 	/* TSO command */
 	HCLGE_OPC_TSO_GENERIC_CONFIG	= 0x0C01,
+	HCLGE_OPC_GRO_GENERIC_CONFIG    = 0x0C10,
 
 	/* RSS commands */
 	HCLGE_OPC_RSS_GENERIC_CONFIG	= 0x0D01,
@@ -758,6 +759,12 @@ struct hclge_cfg_tso_status_cmd {
 	u8 rsv[20];
 };
 
+#define HCLGE_GRO_EN_B		0
+struct hclge_cfg_gro_status_cmd {
+	__le16 gro_en;
+	u8 rsv[22];
+};
+
 #define HCLGE_TSO_MSS_MIN	256
 #define HCLGE_TSO_MSS_MAX	9668
 
diff --git a/drivers/net/ethernet/hisilicon/hns3/hns3pf/hclge_main.c b/drivers/net/ethernet/hisilicon/hns3/hns3pf/hclge_main.c
index 43bfc730a62d..3f8bd11d0824 100644
--- a/drivers/net/ethernet/hisilicon/hns3/hns3pf/hclge_main.c
+++ b/drivers/net/ethernet/hisilicon/hns3/hns3pf/hclge_main.c
@@ -921,6 +921,28 @@ static int hclge_config_tso(struct hclge_dev *hdev, int tso_mss_min,
 	return hclge_cmd_send(&hdev->hw, &desc, 1);
 }
 
+static int hclge_config_gro(struct hclge_dev *hdev, bool en)
+{
+	struct hclge_cfg_gro_status_cmd *req;
+	struct hclge_desc desc;
+	int ret;
+
+	if (!hnae3_dev_gro_supported(hdev))
+		return 0;
+
+	hclge_cmd_setup_basic_desc(&desc, HCLGE_OPC_GRO_GENERIC_CONFIG, false);
+	req = (struct hclge_cfg_gro_status_cmd *)desc.data;
+
+	req->gro_en = cpu_to_le16(en ? 1 : 0);
+
+	ret = hclge_cmd_send(&hdev->hw, &desc, 1);
+	if (ret)
+		dev_err(&hdev->pdev->dev,
+			"GRO hardware config cmd failed, ret = %d\n", ret);
+
+	return ret;
+}
+
 static int hclge_alloc_tqps(struct hclge_dev *hdev)
 {
 	struct hclge_tqp *tqp;
@@ -7090,6 +7112,10 @@ static int hclge_init_ae_dev(struct hnae3_ae_dev *ae_dev)
 		goto err_mdiobus_unreg;
 	}
 
+	ret = hclge_config_gro(hdev, true);
+	if (ret)
+		goto err_mdiobus_unreg;
+
 	ret = hclge_init_vlan_config(hdev);
 	if (ret) {
 		dev_err(&pdev->dev, "VLAN init fail, ret =%d\n", ret);
@@ -7221,6 +7247,10 @@ static int hclge_reset_ae_dev(struct hnae3_ae_dev *ae_dev)
 		return ret;
 	}
 
+	ret = hclge_config_gro(hdev, true);
+	if (ret)
+		return ret;
+
 	ret = hclge_init_vlan_config(hdev);
 	if (ret) {
 		dev_err(&pdev->dev, "VLAN init fail, ret =%d\n", ret);
diff --git a/drivers/net/ethernet/hisilicon/hns3/hns3vf/hclgevf_cmd.h b/drivers/net/ethernet/hisilicon/hns3/hns3vf/hclgevf_cmd.h
index 090541de0c7d..47030b42341f 100644
--- a/drivers/net/ethernet/hisilicon/hns3/hns3vf/hclgevf_cmd.h
+++ b/drivers/net/ethernet/hisilicon/hns3/hns3vf/hclgevf_cmd.h
@@ -87,6 +87,8 @@ enum hclgevf_opcode_type {
 	HCLGEVF_OPC_QUERY_TX_STATUS	= 0x0B03,
 	HCLGEVF_OPC_QUERY_RX_STATUS	= 0x0B13,
 	HCLGEVF_OPC_CFG_COM_TQP_QUEUE	= 0x0B20,
+	/* GRO command */
+	HCLGEVF_OPC_GRO_GENERIC_CONFIG  = 0x0C10,
 	/* RSS cmd */
 	HCLGEVF_OPC_RSS_GENERIC_CONFIG	= 0x0D01,
 	HCLGEVF_OPC_RSS_INPUT_TUPLE     = 0x0D02,
@@ -149,6 +151,12 @@ struct hclgevf_query_res_cmd {
 	__le16 rsv[7];
 };
 
+#define HCLGEVF_GRO_EN_B               0
+struct hclgevf_cfg_gro_status_cmd {
+	__le16 gro_en;
+	u8 rsv[22];
+};
+
 #define HCLGEVF_RSS_DEFAULT_OUTPORT_B	4
 #define HCLGEVF_RSS_HASH_KEY_OFFSET_B	4
 #define HCLGEVF_RSS_HASH_KEY_NUM	16
diff --git a/drivers/net/ethernet/hisilicon/hns3/hns3vf/hclgevf_main.c b/drivers/net/ethernet/hisilicon/hns3/hns3vf/hclgevf_main.c
index 6b4d1477055f..2e889b829976 100644
--- a/drivers/net/ethernet/hisilicon/hns3/hns3vf/hclgevf_main.c
+++ b/drivers/net/ethernet/hisilicon/hns3/hns3vf/hclgevf_main.c
@@ -1655,6 +1655,29 @@ static int hclgevf_init_roce_base_info(struct hclgevf_dev *hdev)
 	return 0;
 }
 
+static int hclgevf_config_gro(struct hclgevf_dev *hdev, bool en)
+{
+	struct hclgevf_cfg_gro_status_cmd *req;
+	struct hclgevf_desc desc;
+	int ret;
+
+	if (!hnae3_dev_gro_supported(hdev))
+		return 0;
+
+	hclgevf_cmd_setup_basic_desc(&desc, HCLGEVF_OPC_GRO_GENERIC_CONFIG,
+				     false);
+	req = (struct hclgevf_cfg_gro_status_cmd *)desc.data;
+
+	req->gro_en = cpu_to_le16(en ? 1 : 0);
+
+	ret = hclgevf_cmd_send(&hdev->hw, &desc, 1);
+	if (ret)
+		dev_err(&hdev->pdev->dev,
+			"VF GRO hardware config cmd failed, ret = %d.\n", ret);
+
+	return ret;
+}
+
 static int hclgevf_rss_init_hw(struct hclgevf_dev *hdev)
 {
 	struct hclgevf_rss_cfg *rss_cfg = &hdev->rss_cfg;
@@ -2122,6 +2145,10 @@ static int hclgevf_reset_hdev(struct hclgevf_dev *hdev)
 		return ret;
 	}
 
+	ret = hclgevf_config_gro(hdev, true);
+	if (ret)
+		return ret;
+
 	ret = hclgevf_init_vlan_config(hdev);
 	if (ret) {
 		dev_err(&hdev->pdev->dev,
@@ -2199,6 +2226,10 @@ static int hclgevf_init_hdev(struct hclgevf_dev *hdev)
 		goto err_config;
 	}
 
+	ret = hclgevf_config_gro(hdev, true);
+	if (ret)
+		goto err_config;
+
 	/* Initialize RSS for this VF */
 	ret = hclgevf_rss_init_hw(hdev);
 	if (ret) {
-- 
2.17.1

[PATCH V2 net-next 2/5] net: hns3: Add handling of GRO Pkts not fully RX'ed in NAPI poll

[Salil Mehta]

From: Peng Li <lipeng321@huawei.com>

The "FE bit" in the description means the last description for
a packets. When HW GRO enable, HW write data to ring every
packet/buffer, there is greater probability that driver handle
with the describtion but HW still not set the "FE bit".

When drier handle the packet and HW still not set "FE bit",
driver stores skb and bd_num in rx ring, and continue to use the
skb and bd_num in next napi.

Signed-off-by: Peng Li <lipeng321@huawei.com>
Signed-off-by: Salil Mehta <salil.mehta@huawei.com>
---
 .../net/ethernet/hisilicon/hns3/hns3_enet.c   | 196 ++++++++++++------
 .../net/ethernet/hisilicon/hns3/hns3_enet.h   |   6 +
 2 files changed, 140 insertions(+), 62 deletions(-)

diff --git a/drivers/net/ethernet/hisilicon/hns3/hns3_enet.c b/drivers/net/ethernet/hisilicon/hns3/hns3_enet.c
index a510ddfd45a5..d8c5e1198670 100644
--- a/drivers/net/ethernet/hisilicon/hns3/hns3_enet.c
+++ b/drivers/net/ethernet/hisilicon/hns3/hns3_enet.c
@@ -2383,6 +2383,90 @@ static bool hns3_parse_vlan_tag(struct hns3_enet_ring *ring,
 	}
 }
 
+static int hns3_alloc_skb(struct hns3_enet_ring *ring, int length,
+			  unsigned char *va)
+{
+#define HNS3_NEED_ADD_FRAG	1
+	struct hns3_desc_cb *desc_cb = &ring->desc_cb[ring->next_to_clean];
+	struct net_device *netdev = ring->tqp->handle->kinfo.netdev;
+	struct sk_buff *skb;
+
+	ring->skb = napi_alloc_skb(&ring->tqp_vector->napi, HNS3_RX_HEAD_SIZE);
+	skb = ring->skb;
+	if (unlikely(!skb)) {
+		netdev_err(netdev, "alloc rx skb fail\n");
+
+		u64_stats_update_begin(&ring->syncp);
+		ring->stats.sw_err_cnt++;
+		u64_stats_update_end(&ring->syncp);
+
+		return -ENOMEM;
+	}
+
+	prefetchw(skb->data);
+
+	ring->pending_buf = 1;
+	if (length <= HNS3_RX_HEAD_SIZE) {
+		memcpy(__skb_put(skb, length), va, ALIGN(length, sizeof(long)));
+
+		/* We can reuse buffer as-is, just make sure it is local */
+		if (likely(page_to_nid(desc_cb->priv) == numa_node_id()))
+			desc_cb->reuse_flag = 1;
+		else /* This page cannot be reused so discard it */
+			put_page(desc_cb->priv);
+
+		ring_ptr_move_fw(ring, next_to_clean);
+		return 0;
+	}
+	u64_stats_update_begin(&ring->syncp);
+	ring->stats.seg_pkt_cnt++;
+	u64_stats_update_end(&ring->syncp);
+
+	ring->pull_len = eth_get_headlen(va, HNS3_RX_HEAD_SIZE);
+	__skb_put(skb, ring->pull_len);
+	hns3_nic_reuse_page(skb, 0, ring, ring->pull_len,
+			    desc_cb);
+	ring_ptr_move_fw(ring, next_to_clean);
+
+	return HNS3_NEED_ADD_FRAG;
+}
+
+static int hns3_add_frag(struct hns3_enet_ring *ring, struct hns3_desc *desc,
+			 struct sk_buff **out_skb, bool pending)
+{
+	struct sk_buff *skb = *out_skb;
+	struct hns3_desc_cb *desc_cb;
+	struct hns3_desc *pre_desc;
+	u32 bd_base_info;
+	int pre_bd;
+
+	/* if there is pending bd, the SW param next_to_clean has moved
+	 * to next and the next is NULL
+	 */
+	if (pending) {
+		pre_bd = (ring->next_to_clean - 1 + ring->desc_num) %
+			ring->desc_num;
+		pre_desc = &ring->desc[pre_bd];
+		bd_base_info = le32_to_cpu(pre_desc->rx.bd_base_info);
+	} else {
+		bd_base_info = le32_to_cpu(desc->rx.bd_base_info);
+	}
+
+	while (!hnae3_get_bit(bd_base_info, HNS3_RXD_FE_B)) {
+		desc = &ring->desc[ring->next_to_clean];
+		desc_cb = &ring->desc_cb[ring->next_to_clean];
+		bd_base_info = le32_to_cpu(desc->rx.bd_base_info);
+		if (!hnae3_get_bit(bd_base_info, HNS3_RXD_VLD_B))
+			return -ENXIO;
+
+		hns3_nic_reuse_page(skb, ring->pending_buf, ring, 0, desc_cb);
+		ring_ptr_move_fw(ring, next_to_clean);
+		ring->pending_buf++;
+	}
+
+	return 0;
+}
+
 static void hns3_set_rx_skb_rss_type(struct hns3_enet_ring *ring,
 				     struct sk_buff *skb)
 {
@@ -2399,18 +2483,16 @@ static void hns3_set_rx_skb_rss_type(struct hns3_enet_ring *ring,
 }
 
 static int hns3_handle_rx_bd(struct hns3_enet_ring *ring,
-			     struct sk_buff **out_skb, int *out_bnum)
+			     struct sk_buff **out_skb)
 {
 	struct net_device *netdev = ring->tqp->handle->kinfo.netdev;
+	struct sk_buff *skb = ring->skb;
 	struct hns3_desc_cb *desc_cb;
 	struct hns3_desc *desc;
-	struct sk_buff *skb;
-	unsigned char *va;
 	u32 bd_base_info;
-	int pull_len;
 	u32 l234info;
 	int length;
-	int bnum;
+	int ret;
 
 	desc = &ring->desc[ring->next_to_clean];
 	desc_cb = &ring->desc_cb[ring->next_to_clean];
@@ -2422,9 +2504,10 @@ static int hns3_handle_rx_bd(struct hns3_enet_ring *ring,
 
 	/* Check valid BD */
 	if (unlikely(!hnae3_get_bit(bd_base_info, HNS3_RXD_VLD_B)))
-		return -EFAULT;
+		return -ENXIO;
 
-	va = (unsigned char *)desc_cb->buf + desc_cb->page_offset;
+	if (!skb)
+		ring->va = (unsigned char *)desc_cb->buf + desc_cb->page_offset;
 
 	/* Prefetch first cache line of first page
 	 * Idea is to cache few bytes of the header of the packet. Our L1 Cache
@@ -2433,62 +2516,42 @@ static int hns3_handle_rx_bd(struct hns3_enet_ring *ring,
 	 * lines. In such a case, single fetch would suffice to cache in the
 	 * relevant part of the header.
 	 */
-	prefetch(va);
+	prefetch(ring->va);
 #if L1_CACHE_BYTES < 128
-	prefetch(va + L1_CACHE_BYTES);
+	prefetch(ring->va + L1_CACHE_BYTES);
 #endif
 
-	skb = *out_skb = napi_alloc_skb(&ring->tqp_vector->napi,
-					HNS3_RX_HEAD_SIZE);
-	if (unlikely(!skb)) {
-		netdev_err(netdev, "alloc rx skb fail\n");
+	if (!skb) {
+		ret = hns3_alloc_skb(ring, length, ring->va);
+		*out_skb = skb = ring->skb;
 
-		u64_stats_update_begin(&ring->syncp);
-		ring->stats.sw_err_cnt++;
-		u64_stats_update_end(&ring->syncp);
-
-		return -ENOMEM;
-	}
-
-	prefetchw(skb->data);
-
-	bnum = 1;
-	if (length <= HNS3_RX_HEAD_SIZE) {
-		memcpy(__skb_put(skb, length), va, ALIGN(length, sizeof(long)));
-
-		/* We can reuse buffer as-is, just make sure it is local */
-		if (likely(page_to_nid(desc_cb->priv) == numa_node_id()))
-			desc_cb->reuse_flag = 1;
-		else /* This page cannot be reused so discard it */
-			put_page(desc_cb->priv);
+		if (ret < 0) /* alloc buffer fail */
+			return ret;
+		if (ret > 0) { /* need add frag */
+			ret = hns3_add_frag(ring, desc, &skb, false);
+			if (ret)
+				return ret;
 
-		ring_ptr_move_fw(ring, next_to_clean);
+			/* As the head data may be changed when GRO enable, copy
+			 * the head data in after other data rx completed
+			 */
+			memcpy(skb->data, ring->va,
+			       ALIGN(ring->pull_len, sizeof(long)));
+		}
 	} else {
-		u64_stats_update_begin(&ring->syncp);
-		ring->stats.seg_pkt_cnt++;
-		u64_stats_update_end(&ring->syncp);
-
-		pull_len = eth_get_headlen(va, HNS3_RX_HEAD_SIZE);
-
-		memcpy(__skb_put(skb, pull_len), va,
-		       ALIGN(pull_len, sizeof(long)));
-
-		hns3_nic_reuse_page(skb, 0, ring, pull_len, desc_cb);
-		ring_ptr_move_fw(ring, next_to_clean);
+		ret = hns3_add_frag(ring, desc, &skb, true);
+		if (ret)
+			return ret;
 
-		while (!hnae3_get_bit(bd_base_info, HNS3_RXD_FE_B)) {
-			desc = &ring->desc[ring->next_to_clean];
-			desc_cb = &ring->desc_cb[ring->next_to_clean];
-			bd_base_info = le32_to_cpu(desc->rx.bd_base_info);
-			hns3_nic_reuse_page(skb, bnum, ring, 0, desc_cb);
-			ring_ptr_move_fw(ring, next_to_clean);
-			bnum++;
-		}
+		/* As the head data may be changed when GRO enable, copy
+		 * the head data in after other data rx completed
+		 */
+		memcpy(skb->data, ring->va,
+		       ALIGN(ring->pull_len, sizeof(long)));
 	}
 
-	*out_bnum = bnum;
-
 	l234info = le32_to_cpu(desc->rx.l234_info);
+	bd_base_info = le32_to_cpu(desc->rx.bd_base_info);
 
 	/* Based on hw strategy, the tag offloaded will be stored at
 	 * ot_vlan_tag in two layer tag case, and stored at vlan_tag
@@ -2539,6 +2602,7 @@ static int hns3_handle_rx_bd(struct hns3_enet_ring *ring,
 	ring->tqp_vector->rx_group.total_bytes += skb->len;
 
 	hns3_rx_checksum(ring, skb, desc);
+	*out_skb = skb;
 	hns3_set_rx_skb_rss_type(ring, skb);
 
 	return 0;
@@ -2551,9 +2615,9 @@ int hns3_clean_rx_ring(
 #define RCB_NOF_ALLOC_RX_BUFF_ONCE 16
 	struct net_device *netdev = ring->tqp->handle->kinfo.netdev;
 	int recv_pkts, recv_bds, clean_count, err;
-	int unused_count = hns3_desc_unused(ring);
-	struct sk_buff *skb = NULL;
-	int num, bnum = 0;
+	int unused_count = hns3_desc_unused(ring) - ring->pending_buf;
+	struct sk_buff *skb = ring->skb;
+	int num;
 
 	num = readl_relaxed(ring->tqp->io_base + HNS3_RING_RX_RING_FBDNUM_REG);
 	rmb(); /* Make sure num taken effect before the other data is touched */
@@ -2567,24 +2631,32 @@ int hns3_clean_rx_ring(
 			hns3_nic_alloc_rx_buffers(ring,
 						  clean_count + unused_count);
 			clean_count = 0;
-			unused_count = hns3_desc_unused(ring);
+			unused_count = hns3_desc_unused(ring) -
+					ring->pending_buf;
 		}
 
 		/* Poll one pkt */
-		err = hns3_handle_rx_bd(ring, &skb, &bnum);
+		err = hns3_handle_rx_bd(ring, &skb);
 		if (unlikely(!skb)) /* This fault cannot be repaired */
 			goto out;
 
-		recv_bds += bnum;
-		clean_count += bnum;
-		if (unlikely(err)) {  /* Do jump the err */
-			recv_pkts++;
+		if (err == -ENXIO) { /* Do not get FE for the packet */
+			goto out;
+		} else if (unlikely(err)) {  /* Do jump the err */
+			recv_bds += ring->pending_buf;
+			clean_count += ring->pending_buf;
+			ring->skb = NULL;
+			ring->pending_buf = 0;
 			continue;
 		}
 
 		/* Do update ip stack process */
 		skb->protocol = eth_type_trans(skb, netdev);
 		rx_fn(ring, skb);
+		recv_bds += ring->pending_buf;
+		clean_count += ring->pending_buf;
+		ring->skb = NULL;
+		ring->pending_buf = 0;
 
 		recv_pkts++;
 	}
diff --git a/drivers/net/ethernet/hisilicon/hns3/hns3_enet.h b/drivers/net/ethernet/hisilicon/hns3/hns3_enet.h
index 10ff18af3cc7..d8c0998127be 100644
--- a/drivers/net/ethernet/hisilicon/hns3/hns3_enet.h
+++ b/drivers/net/ethernet/hisilicon/hns3/hns3_enet.h
@@ -401,11 +401,17 @@ struct hns3_enet_ring {
 	 */
 	int next_to_clean;
 
+	int pull_len; /* head length for current packet */
+	unsigned char *va; /* first buffer address for current packet */
+
 	u32 flag;          /* ring attribute */
 	int irq_init_flag;
 
 	int numa_node;
 	cpumask_t affinity_mask;
+
+	int pending_buf;
+	struct sk_buff *skb;
 };
 
 struct hns_queue;
-- 
2.17.1

[PATCH V2 net-next 3/5] net: hns3: Add support for ethtool -K to enable/disable HW GRO

[Salil Mehta]

From: Peng Li <lipeng321@huawei.com>

This patch adds support of ethtool -K to enable/disable
hardware GRO in HNS3 PF/VF driver.

Signed-off-by: Peng Li <lipeng321@huawei.com>
Signed-off-by: Salil Mehta <salil.mehta@huawei.com>
---
 drivers/net/ethernet/hisilicon/hns3/hnae3.h         |  3 +++
 drivers/net/ethernet/hisilicon/hns3/hns3_enet.c     | 13 ++++++++++++-
 .../net/ethernet/hisilicon/hns3/hns3pf/hclge_main.c |  9 +++++++++
 .../ethernet/hisilicon/hns3/hns3vf/hclgevf_main.c   |  8 ++++++++
 4 files changed, 32 insertions(+), 1 deletion(-)

diff --git a/drivers/net/ethernet/hisilicon/hns3/hnae3.h b/drivers/net/ethernet/hisilicon/hns3/hnae3.h
index 21d934b7a2a3..1746172ffbf7 100644
--- a/drivers/net/ethernet/hisilicon/hns3/hnae3.h
+++ b/drivers/net/ethernet/hisilicon/hns3/hnae3.h
@@ -305,6 +305,8 @@ struct hnae3_ae_dev {
  *   Set vlan filter config of vf
  * enable_hw_strip_rxvtag()
  *   Enable/disable hardware strip vlan tag of packets received
+ * set_gro_en
+ *   Enable/disable HW GRO
  */
 struct hnae3_ae_ops {
 	int (*init_ae_dev)(struct hnae3_ae_dev *ae_dev);
@@ -449,6 +451,7 @@ struct hnae3_ae_ops {
 	bool (*get_hw_reset_stat)(struct hnae3_handle *handle);
 	bool (*ae_dev_resetting)(struct hnae3_handle *handle);
 	unsigned long (*ae_dev_reset_cnt)(struct hnae3_handle *handle);
+	int (*set_gro_en)(struct hnae3_handle *handle, int enable);
 };
 
 struct hnae3_dcb_ops {
diff --git a/drivers/net/ethernet/hisilicon/hns3/hns3_enet.c b/drivers/net/ethernet/hisilicon/hns3/hns3_enet.c
index d8c5e1198670..860067898471 100644
--- a/drivers/net/ethernet/hisilicon/hns3/hns3_enet.c
+++ b/drivers/net/ethernet/hisilicon/hns3/hns3_enet.c
@@ -1345,6 +1345,15 @@ static int hns3_nic_set_features(struct net_device *netdev,
 			priv->ops.maybe_stop_tx = hns3_nic_maybe_stop_tx;
 	}
 
+	if (changed & (NETIF_F_GRO_HW) && h->ae_algo->ops->set_gro_en) {
+		if (features & NETIF_F_GRO_HW)
+			ret = h->ae_algo->ops->set_gro_en(h, true);
+		else
+			ret = h->ae_algo->ops->set_gro_en(h, false);
+		if (ret)
+			return ret;
+	}
+
 	if ((changed & NETIF_F_HW_VLAN_CTAG_FILTER) &&
 	    h->ae_algo->ops->enable_vlan_filter) {
 		if (features & NETIF_F_HW_VLAN_CTAG_FILTER)
@@ -1929,7 +1938,9 @@ static void hns3_set_default_feature(struct net_device *netdev)
 		NETIF_F_GSO_UDP_TUNNEL_CSUM | NETIF_F_SCTP_CRC;
 
 	if (pdev->revision >= 0x21) {
-		netdev->hw_features |= NETIF_F_HW_VLAN_CTAG_FILTER;
+		netdev->hw_features |= NETIF_F_HW_VLAN_CTAG_FILTER |
+			NETIF_F_GRO_HW;
+		netdev->features |= NETIF_F_GRO_HW;
 
 		if (!(h->flags & HNAE3_SUPPORT_VF)) {
 			netdev->hw_features |= NETIF_F_NTUPLE;
diff --git a/drivers/net/ethernet/hisilicon/hns3/hns3pf/hclge_main.c b/drivers/net/ethernet/hisilicon/hns3/hns3pf/hclge_main.c
index 3f8bd11d0824..eb6b6a5f744f 100644
--- a/drivers/net/ethernet/hisilicon/hns3/hns3pf/hclge_main.c
+++ b/drivers/net/ethernet/hisilicon/hns3/hns3pf/hclge_main.c
@@ -7667,6 +7667,14 @@ static void hclge_get_link_mode(struct hnae3_handle *handle,
 	}
 }
 
+static int hclge_gro_en(struct hnae3_handle *handle, int enable)
+{
+	struct hclge_vport *vport = hclge_get_vport(handle);
+	struct hclge_dev *hdev = vport->back;
+
+	return hclge_config_gro(hdev, enable);
+}
+
 static const struct hnae3_ae_ops hclge_ops = {
 	.init_ae_dev = hclge_init_ae_dev,
 	.uninit_ae_dev = hclge_uninit_ae_dev,
@@ -7738,6 +7746,7 @@ static const struct hnae3_ae_ops hclge_ops = {
 	.get_hw_reset_stat = hclge_get_hw_reset_stat,
 	.ae_dev_resetting = hclge_ae_dev_resetting,
 	.ae_dev_reset_cnt = hclge_ae_dev_reset_cnt,
+	.set_gro_en = hclge_gro_en,
 };
 
 static struct hnae3_ae_algo ae_algo = {
diff --git a/drivers/net/ethernet/hisilicon/hns3/hns3vf/hclgevf_main.c b/drivers/net/ethernet/hisilicon/hns3/hns3vf/hclgevf_main.c
index 2e889b829976..e27d6fa85a74 100644
--- a/drivers/net/ethernet/hisilicon/hns3/hns3vf/hclgevf_main.c
+++ b/drivers/net/ethernet/hisilicon/hns3/hns3vf/hclgevf_main.c
@@ -2368,6 +2368,13 @@ void hclgevf_update_speed_duplex(struct hclgevf_dev *hdev, u32 speed,
 	hdev->hw.mac.duplex = duplex;
 }
 
+static int hclgevf_gro_en(struct hnae3_handle *handle, int enable)
+{
+	struct hclgevf_dev *hdev = hclgevf_ae_get_hdev(handle);
+
+	return hclgevf_config_gro(hdev, enable);
+}
+
 static void hclgevf_get_media_type(struct hnae3_handle *handle,
 				  u8 *media_type)
 {
@@ -2442,6 +2449,7 @@ static const struct hnae3_ae_ops hclgevf_ops = {
 	.get_hw_reset_stat = hclgevf_get_hw_reset_stat,
 	.ae_dev_resetting = hclgevf_ae_dev_resetting,
 	.ae_dev_reset_cnt = hclgevf_ae_dev_reset_cnt,
+	.set_gro_en = hclgevf_gro_en,
 };
 
 static struct hnae3_ae_algo ae_algovf = {
-- 
2.17.1

[PATCH V2 net-next 4/5] net: hns3: Add skb chain when num of RX buf exceeds MAX_SKB_FRAGS

[Salil Mehta]

From: Peng Li <lipeng321@huawei.com>

MAX_SKB_FRAGS in protocol stack is defined as:

MAX_SKB_FRAGS is 17 when PAGE_SIZE is 4K. If HW enable GRO, it may
merge small packets and the rx buffer may be more than
MAX_SKB_FRAGS. So driver will add skb chain when RX buffer num.
more than MAX_SKB_FRAGS.

Signed-off-by: Peng Li <lipeng321@huawei.com>
Signed-off-by: Salil Mehta <salil.mehta@huawei.com>
---
 .../net/ethernet/hisilicon/hns3/hns3_enet.c   | 37 ++++++++++++++++++-
 .../net/ethernet/hisilicon/hns3/hns3_enet.h   |  2 +
 2 files changed, 37 insertions(+), 2 deletions(-)

diff --git a/drivers/net/ethernet/hisilicon/hns3/hns3_enet.c b/drivers/net/ethernet/hisilicon/hns3/hns3_enet.c
index 860067898471..7776089b6bc2 100644
--- a/drivers/net/ethernet/hisilicon/hns3/hns3_enet.c
+++ b/drivers/net/ethernet/hisilicon/hns3/hns3_enet.c
@@ -2361,6 +2361,9 @@ static void hns3_rx_checksum(struct hns3_enet_ring *ring, struct sk_buff *skb,
 
 static void hns3_rx_skb(struct hns3_enet_ring *ring, struct sk_buff *skb)
 {
+	if (skb_has_frag_list(skb))
+		napi_gro_flush(&ring->tqp_vector->napi, false);
+
 	napi_gro_receive(&ring->tqp_vector->napi, skb);
 }
 
@@ -2417,6 +2420,8 @@ static int hns3_alloc_skb(struct hns3_enet_ring *ring, int length,
 	prefetchw(skb->data);
 
 	ring->pending_buf = 1;
+	ring->frag_num = 0;
+	ring->tail_skb = NULL;
 	if (length <= HNS3_RX_HEAD_SIZE) {
 		memcpy(__skb_put(skb, length), va, ALIGN(length, sizeof(long)));
 
@@ -2435,7 +2440,7 @@ static int hns3_alloc_skb(struct hns3_enet_ring *ring, int length,
 
 	ring->pull_len = eth_get_headlen(va, HNS3_RX_HEAD_SIZE);
 	__skb_put(skb, ring->pull_len);
-	hns3_nic_reuse_page(skb, 0, ring, ring->pull_len,
+	hns3_nic_reuse_page(skb, ring->frag_num++, ring, ring->pull_len,
 			    desc_cb);
 	ring_ptr_move_fw(ring, next_to_clean);
 
@@ -2446,6 +2451,8 @@ static int hns3_add_frag(struct hns3_enet_ring *ring, struct hns3_desc *desc,
 			 struct sk_buff **out_skb, bool pending)
 {
 	struct sk_buff *skb = *out_skb;
+	struct sk_buff *head_skb = *out_skb;
+	struct sk_buff *new_skb;
 	struct hns3_desc_cb *desc_cb;
 	struct hns3_desc *pre_desc;
 	u32 bd_base_info;
@@ -2470,7 +2477,33 @@ static int hns3_add_frag(struct hns3_enet_ring *ring, struct hns3_desc *desc,
 		if (!hnae3_get_bit(bd_base_info, HNS3_RXD_VLD_B))
 			return -ENXIO;
 
-		hns3_nic_reuse_page(skb, ring->pending_buf, ring, 0, desc_cb);
+		if (unlikely(ring->frag_num >= MAX_SKB_FRAGS)) {
+			new_skb = napi_alloc_skb(&ring->tqp_vector->napi,
+						 HNS3_RX_HEAD_SIZE);
+			if (unlikely(!new_skb)) {
+				netdev_err(ring->tqp->handle->kinfo.netdev,
+					   "alloc rx skb frag fail\n");
+				return -ENXIO;
+			}
+			ring->frag_num = 0;
+
+			if (ring->tail_skb) {
+				ring->tail_skb->next = new_skb;
+				ring->tail_skb = new_skb;
+			} else {
+				skb_shinfo(skb)->frag_list = new_skb;
+				ring->tail_skb = new_skb;
+			}
+		}
+
+		if (ring->tail_skb) {
+			head_skb->truesize += hnae3_buf_size(ring);
+			head_skb->data_len += le16_to_cpu(desc->rx.size);
+			head_skb->len += le16_to_cpu(desc->rx.size);
+			skb = ring->tail_skb;
+		}
+
+		hns3_nic_reuse_page(skb, ring->frag_num++, ring, 0, desc_cb);
 		ring_ptr_move_fw(ring, next_to_clean);
 		ring->pending_buf++;
 	}
diff --git a/drivers/net/ethernet/hisilicon/hns3/hns3_enet.h b/drivers/net/ethernet/hisilicon/hns3/hns3_enet.h
index d8c0998127be..8e56b7e44978 100644
--- a/drivers/net/ethernet/hisilicon/hns3/hns3_enet.h
+++ b/drivers/net/ethernet/hisilicon/hns3/hns3_enet.h
@@ -402,6 +402,7 @@ struct hns3_enet_ring {
 	int next_to_clean;
 
 	int pull_len; /* head length for current packet */
+	u32 frag_num;
 	unsigned char *va; /* first buffer address for current packet */
 
 	u32 flag;          /* ring attribute */
@@ -412,6 +413,7 @@ struct hns3_enet_ring {
 
 	int pending_buf;
 	struct sk_buff *skb;
+	struct sk_buff *tail_skb;
 };
 
 struct hns_queue;
-- 
2.17.1

[PATCH V2 net-next 5/5] net: hns3: Adds GRO params to SKB for the stack

[Salil Mehta]

From: Peng Li <lipeng321@huawei.com>

When HW GRO enable, protocol stack will not do GRO again,
driver should add gro param to the skb for the protocol
stack..

Signed-off-by: Peng Li <lipeng321@huawei.com>
Signed-off-by: Salil Mehta <salil.mehta@huawei.com>
---
 .../net/ethernet/hisilicon/hns3/hns3_enet.c   | 43 +++++++++++++++++++
 .../net/ethernet/hisilicon/hns3/hns3_enet.h   |  9 ++--
 2 files changed, 49 insertions(+), 3 deletions(-)

diff --git a/drivers/net/ethernet/hisilicon/hns3/hns3_enet.c b/drivers/net/ethernet/hisilicon/hns3/hns3_enet.c
index 7776089b6bc2..22220af92aa9 100644
--- a/drivers/net/ethernet/hisilicon/hns3/hns3_enet.c
+++ b/drivers/net/ethernet/hisilicon/hns3/hns3_enet.c
@@ -15,6 +15,7 @@
 #include <linux/vermagic.h>
 #include <net/gre.h>
 #include <net/pkt_cls.h>
+#include <net/tcp.h>
 #include <net/vxlan.h>
 
 #include "hnae3.h"
@@ -2318,6 +2319,12 @@ static void hns3_rx_checksum(struct hns3_enet_ring *ring, struct sk_buff *skb,
 	if (!(netdev->features & NETIF_F_RXCSUM))
 		return;
 
+	/* We MUST enable hardware checksum before enabling hardware GRO */
+	if (skb_shinfo(skb)->gso_size) {
+		skb->ip_summed = CHECKSUM_UNNECESSARY;
+		return;
+	}
+
 	/* check if hardware has done checksum */
 	if (!hnae3_get_bit(bd_base_info, HNS3_RXD_L3L4P_B))
 		return;
@@ -2511,6 +2518,39 @@ static int hns3_add_frag(struct hns3_enet_ring *ring, struct hns3_desc *desc,
 	return 0;
 }
 
+static void hns3_set_gro_param(struct sk_buff *skb, u32 l234info,
+			       u32 bd_base_info)
+{
+	u16 gro_count;
+	u32 l3_type;
+
+	gro_count = hnae3_get_field(l234info, HNS3_RXD_GRO_COUNT_M,
+				    HNS3_RXD_GRO_COUNT_S);
+	/* if there is no HW GRO, do not set gro params */
+	if (!gro_count)
+		return;
+
+	/* tcp_gro_complete() will copy NAPI_GRO_CB(skb)->count
+	 * to skb_shinfo(skb)->gso_segs
+	 */
+	NAPI_GRO_CB(skb)->count = gro_count;
+
+	l3_type = hnae3_get_field(l234info, HNS3_RXD_L3ID_M,
+				  HNS3_RXD_L3ID_S);
+	if (l3_type == HNS3_L3_TYPE_IPV4)
+		skb_shinfo(skb)->gso_type = SKB_GSO_TCPV4;
+	else if (l3_type == HNS3_L3_TYPE_IPV6)
+		skb_shinfo(skb)->gso_type = SKB_GSO_TCPV6;
+	else
+		return;
+
+	skb_shinfo(skb)->gso_size = hnae3_get_field(bd_base_info,
+						    HNS3_RXD_GRO_SIZE_M,
+						    HNS3_RXD_GRO_SIZE_S);
+	if (skb_shinfo(skb)->gso_size)
+		tcp_gro_complete(skb);
+}
+
 static void hns3_set_rx_skb_rss_type(struct hns3_enet_ring *ring,
 				     struct sk_buff *skb)
 {
@@ -2645,6 +2685,9 @@ static int hns3_handle_rx_bd(struct hns3_enet_ring *ring,
 
 	ring->tqp_vector->rx_group.total_bytes += skb->len;
 
+	/* This is needed in order to enable forwarding support */
+	hns3_set_gro_param(skb, l234info, bd_base_info);
+
 	hns3_rx_checksum(ring, skb, desc);
 	*out_skb = skb;
 	hns3_set_rx_skb_rss_type(ring, skb);
diff --git a/drivers/net/ethernet/hisilicon/hns3/hns3_enet.h b/drivers/net/ethernet/hisilicon/hns3/hns3_enet.h
index 8e56b7e44978..3365c9596983 100644
--- a/drivers/net/ethernet/hisilicon/hns3/hns3_enet.h
+++ b/drivers/net/ethernet/hisilicon/hns3/hns3_enet.h
@@ -109,6 +109,10 @@ enum hns3_nic_state {
 #define HNS3_RXD_DOI_B				21
 #define HNS3_RXD_OL3E_B				22
 #define HNS3_RXD_OL4E_B				23
+#define HNS3_RXD_GRO_COUNT_S			24
+#define HNS3_RXD_GRO_COUNT_M			(0x3f << HNS3_RXD_GRO_COUNT_S)
+#define HNS3_RXD_GRO_FIXID_B			30
+#define HNS3_RXD_GRO_ECN_B			31
 
 #define HNS3_RXD_ODMAC_S			0
 #define HNS3_RXD_ODMAC_M			(0x3 << HNS3_RXD_ODMAC_S)
@@ -135,9 +139,8 @@ enum hns3_nic_state {
 #define HNS3_RXD_TSIND_S			12
 #define HNS3_RXD_TSIND_M			(0x7 << HNS3_RXD_TSIND_S)
 #define HNS3_RXD_LKBK_B				15
-#define HNS3_RXD_HDL_S				16
-#define HNS3_RXD_HDL_M				(0x7ff << HNS3_RXD_HDL_S)
-#define HNS3_RXD_HSIND_B			31
+#define HNS3_RXD_GRO_SIZE_S			16
+#define HNS3_RXD_GRO_SIZE_M			(0x3ff << HNS3_RXD_GRO_SIZE_S)
 
 #define HNS3_TXD_L3T_S				0
 #define HNS3_TXD_L3T_M				(0x3 << HNS3_TXD_L3T_S)
-- 
2.17.1

Re: [PATCH V2 net-next 0/5] net: hns3: Add support of hardware GRO to HNS3 Driver

[David Miller]

From: Salil Mehta <salil.mehta@huawei.com>
Date: Thu, 15 Nov 2018 09:29:20 +0000

> This patch-set adds support of hardware assisted GRO feature to
> HNS3 driver on Rev B(=0x21) platform. Current hardware only
> supports TCP/IPv{4|6} flows.
> 
> Change Log:
> V1->V2:
> 1. Remove redundant print reported by Leon Romanovsky.
>    Link: https://lkml.org/lkml/2018/11/13/715

Series applied, thanks.