net: hns3: use tx bounce buffer for small packets

when the packet or frag size is small, it causes both security and
performance issue. As dma can't map sub-page, this means some extra
kernel data is visible to devices. On the other hand, the overhead
of dma map and unmap is huge when IOMMU is on.

So add a queue based tx shared bounce buffer to memcpy the small
packet when the len of the xmitted skb is below tx_copybreak.
Add tx_spare_buf_size module param to set the size of tx spare
buffer, and add set/get_tunable to set or query the tx_copybreak.

The throughtput improves from 30 Gbps to 90+ Gbps when running 16
netperf threads with 32KB UDP message size when IOMMU is in the
strict mode(tx_copybreak = 2000 and mtu = 1500).

Suggested-by: Barry Song <song.bao.hua@hisilicon.com>
Signed-off-by: Yunsheng Lin <linyunsheng@huawei.com>
Signed-off-by: Guangbin Huang <huangguangbin2@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

1 files changed, 34 insertions, 9 deletions

diff --git a/drivers/net/ethernet/hisilicon/hns3/hns3_enet.h b/drivers/net/ethernet/hisilicon/hns3/hns3_enet.h
index 9d18b9430b54..8d147c1dab2c 100644
--- a/drivers/net/ethernet/hisilicon/hns3/hns3_enet.h
+++ b/drivers/net/ethernet/hisilicon/hns3/hns3_enet.h
@@ -304,6 +304,8 @@ enum hns3_desc_type {
 	DESC_TYPE_SKB			= 1 << 0,
 	DESC_TYPE_FRAGLIST_SKB		= 1 << 1,
 	DESC_TYPE_PAGE			= 1 << 2,
+	DESC_TYPE_BOUNCE_ALL		= 1 << 3,
+	DESC_TYPE_BOUNCE_HEAD		= 1 << 4,
 };
 
 struct hns3_desc_cb {
@@ -405,6 +407,9 @@ struct ring_stats {
 			u64 tx_tso_err;
 			u64 over_max_recursion;
 			u64 hw_limitation;
+			u64 tx_bounce;
+			u64 tx_spare_full;
+			u64 copy_bits_err;
 		};
 		struct {
 			u64 rx_pkts;
@@ -423,6 +428,15 @@ struct ring_stats {
 	};
 };
 
+struct hns3_tx_spare {
+	dma_addr_t dma;
+	void *buf;
+	u32 next_to_use;
+	u32 next_to_clean;
+	u32 last_to_clean;
+	u32 len;
+};
+
 struct hns3_enet_ring {
 	struct hns3_desc *desc; /* dma map address space */
 	struct hns3_desc_cb *desc_cb;
@@ -445,18 +459,28 @@ struct hns3_enet_ring {
 	 * next_to_use
 	 */
 	int next_to_clean;
-	union {
-		int last_to_use;	/* last idx used by xmit */
-		u32 pull_len;		/* memcpy len for current rx packet */
-	};
-	u32 frag_num;
-	void *va; /* first buffer address for current packet */
-
 	u32 flag;          /* ring attribute */
 
 	int pending_buf;
-	struct sk_buff *skb;
-	struct sk_buff *tail_skb;
+	union {
+		/* for Tx ring */
+		struct {
+			u32 fd_qb_tx_sample;
+			int last_to_use;        /* last idx used by xmit */
+			u32 tx_copybreak;
+			struct hns3_tx_spare *tx_spare;
+		};
+
+		/* for Rx ring */
+		struct {
+			u32 pull_len;   /* memcpy len for current rx packet */
+			u32 frag_num;
+			/* first buffer address for current packet */
+			unsigned char *va;
+			struct sk_buff *skb;
+			struct sk_buff *tail_skb;
+		};
+	};
 } ____cacheline_internodealigned_in_smp;
 
 enum hns3_flow_level_range {
@@ -540,6 +564,7 @@ struct hns3_nic_priv {
 
 	struct hns3_enet_coalesce tx_coal;
 	struct hns3_enet_coalesce rx_coal;
+	u32 tx_copybreak;
 };
 
 union l3_hdr_info {