Merge branch 'net-ethool-add-support-to-get-set-tx-push-by-ethtool-g-g'

Jie Wang says:

====================
net: ethool: add support to get/set tx push by ethtool -G/g

These three patches add tx push in ring params and adapt the set and get APIs
of ring params.
====================

Link: https://lore.kernel.org/r/20220412020121.14140-1-huangguangbin2@huawei.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

6 files changed, 80 insertions, 22 deletions

diff --git a/Documentation/networking/ethtool-netlink.rst b/Documentation/networking/ethtool-netlink.rst
index 24d9be69065d3c..dbca3e9ec782f6 100644
--- a/Documentation/networking/ethtool-netlink.rst
+++ b/Documentation/networking/ethtool-netlink.rst
@@ -862,6 +862,7 @@ Kernel response contents:
   ``ETHTOOL_A_RINGS_RX_BUF_LEN``        u32     size of buffers on the ring
   ``ETHTOOL_A_RINGS_TCP_DATA_SPLIT``    u8      TCP header / data split
   ``ETHTOOL_A_RINGS_CQE_SIZE``          u32     Size of TX/RX CQE
+  ``ETHTOOL_A_RINGS_TX_PUSH``           u8      flag of TX Push mode
   ====================================  ======  ===========================
 
 ``ETHTOOL_A_RINGS_TCP_DATA_SPLIT`` indicates whether the device is usable with
@@ -871,6 +872,12 @@ separate buffers. The device configuration must make it possible to receive
 full memory pages of data, for example because MTU is high enough or through
 HW-GRO.
 
+``ETHTOOL_A_RINGS_TX_PUSH`` flag is used to enable descriptor fast
+path to send packets. In ordinary path, driver fills descriptors in DRAM and
+notifies NIC hardware. In fast path, driver pushes descriptors to the device
+through MMIO writes, thus reducing the latency. However, enabling this feature
+may increase the CPU cost. Drivers may enforce additional per-packet
+eligibility checks (e.g. on packet size).
 
 RINGS_SET
 =========
@@ -887,6 +894,7 @@ Request contents:
   ``ETHTOOL_A_RINGS_TX``                u32     size of TX ring
   ``ETHTOOL_A_RINGS_RX_BUF_LEN``        u32     size of buffers on the ring
   ``ETHTOOL_A_RINGS_CQE_SIZE``          u32     Size of TX/RX CQE
+  ``ETHTOOL_A_RINGS_TX_PUSH``           u8      flag of TX Push mode
   ====================================  ======  ===========================
 
 Kernel checks that requested ring sizes do not exceed limits reported by
diff --git a/drivers/net/ethernet/hisilicon/hns3/hns3_ethtool.c b/drivers/net/ethernet/hisilicon/hns3/hns3_ethtool.c
index f4da77452126f8..9f4111fd298681 100644
--- a/drivers/net/ethernet/hisilicon/hns3/hns3_ethtool.c
+++ b/drivers/net/ethernet/hisilicon/hns3/hns3_ethtool.c
@@ -664,6 +664,8 @@ static void hns3_get_ringparam(struct net_device *netdev,
 	param->tx_pending = priv->ring[0].desc_num;
 	param->rx_pending = priv->ring[rx_queue_index].desc_num;
 	kernel_param->rx_buf_len = priv->ring[rx_queue_index].buf_size;
+	kernel_param->tx_push = test_bit(HNS3_NIC_STATE_TX_PUSH_ENABLE,
+					 &priv->state);
 }
 
 static void hns3_get_pauseparam(struct net_device *netdev,
@@ -1120,6 +1122,30 @@ static int hns3_change_rx_buf_len(struct net_device *ndev, u32 rx_buf_len)
 	return 0;
 }
 
+static int hns3_set_tx_push(struct net_device *netdev, u32 tx_push)
+{
+	struct hns3_nic_priv *priv = netdev_priv(netdev);
+	struct hnae3_handle *h = hns3_get_handle(netdev);
+	struct hnae3_ae_dev *ae_dev = pci_get_drvdata(h->pdev);
+	u32 old_state = test_bit(HNS3_NIC_STATE_TX_PUSH_ENABLE, &priv->state);
+
+	if (!test_bit(HNAE3_DEV_SUPPORT_TX_PUSH_B, ae_dev->caps) && tx_push)
+		return -EOPNOTSUPP;
+
+	if (tx_push == old_state)
+		return 0;
+
+	netdev_dbg(netdev, "Changing tx push from %s to %s\n",
+		   old_state ? "on" : "off", tx_push ? "on" : "off");
+
+	if (tx_push)
+		set_bit(HNS3_NIC_STATE_TX_PUSH_ENABLE, &priv->state);
+	else
+		clear_bit(HNS3_NIC_STATE_TX_PUSH_ENABLE, &priv->state);
+
+	return 0;
+}
+
 static int hns3_set_ringparam(struct net_device *ndev,
 			      struct ethtool_ringparam *param,
 			      struct kernel_ethtool_ringparam *kernel_param,
@@ -1139,6 +1165,10 @@ static int hns3_set_ringparam(struct net_device *ndev,
 	if (ret)
 		return ret;
 
+	ret = hns3_set_tx_push(ndev, kernel_param->tx_push);
+	if (ret)
+		return ret;
+
 	/* Hardware requires that its descriptors must be multiple of eight */
 	new_tx_desc_num = ALIGN(param->tx_pending, HNS3_RING_BD_MULTIPLE);
 	new_rx_desc_num = ALIGN(param->rx_pending, HNS3_RING_BD_MULTIPLE);
@@ -1858,7 +1888,8 @@ static int hns3_set_tunable(struct net_device *netdev,
 				 ETHTOOL_COALESCE_MAX_FRAMES |		\
 				 ETHTOOL_COALESCE_USE_CQE)
 
-#define HNS3_ETHTOOL_RING	ETHTOOL_RING_USE_RX_BUF_LEN
+#define HNS3_ETHTOOL_RING	(ETHTOOL_RING_USE_RX_BUF_LEN |		\
+				 ETHTOOL_RING_USE_TX_PUSH)
 
 static int hns3_get_ts_info(struct net_device *netdev,
 			    struct ethtool_ts_info *info)
diff --git a/include/linux/ethtool.h b/include/linux/ethtool.h
index 4af58459a1e79c..99dc7bfbcd3c32 100644
--- a/include/linux/ethtool.h
+++ b/include/linux/ethtool.h
@@ -71,11 +71,13 @@ enum {
  * struct kernel_ethtool_ringparam - RX/TX ring configuration
  * @rx_buf_len: Current length of buffers on the rx ring.
  * @tcp_data_split: Scatter packet headers and data to separate buffers
+ * @tx_push: The flag of tx push mode
  * @cqe_size: Size of TX/RX completion queue event
  */
 struct kernel_ethtool_ringparam {
 	u32	rx_buf_len;
 	u8	tcp_data_split;
+	u8	tx_push;
 	u32	cqe_size;
 };
 
@@ -83,10 +85,12 @@ struct kernel_ethtool_ringparam {
  * enum ethtool_supported_ring_param - indicator caps for setting ring params
  * @ETHTOOL_RING_USE_RX_BUF_LEN: capture for setting rx_buf_len
  * @ETHTOOL_RING_USE_CQE_SIZE: capture for setting cqe_size
+ * @ETHTOOL_RING_USE_TX_PUSH: capture for setting tx_push
  */
 enum ethtool_supported_ring_param {
 	ETHTOOL_RING_USE_RX_BUF_LEN = BIT(0),
 	ETHTOOL_RING_USE_CQE_SIZE   = BIT(1),
+	ETHTOOL_RING_USE_TX_PUSH    = BIT(2),
 };
 
 #define __ETH_RSS_HASH_BIT(bit)	((u32)1 << (bit))
diff --git a/include/uapi/linux/ethtool_netlink.h b/include/uapi/linux/ethtool_netlink.h
index 979850221b8dd8..d2fb4f7be61b2a 100644
--- a/include/uapi/linux/ethtool_netlink.h
+++ b/include/uapi/linux/ethtool_netlink.h
@@ -338,6 +338,7 @@ enum {
 	ETHTOOL_A_RINGS_RX_BUF_LEN,                     /* u32 */
 	ETHTOOL_A_RINGS_TCP_DATA_SPLIT,			/* u8 */
 	ETHTOOL_A_RINGS_CQE_SIZE,			/* u32 */
+	ETHTOOL_A_RINGS_TX_PUSH,			/* u8 */
 
 	/* add new constants above here */
 	__ETHTOOL_A_RINGS_CNT,
diff --git a/net/ethtool/netlink.h b/net/ethtool/netlink.h
index 29d01662a48ba1..7919ddb2371ce0 100644
--- a/net/ethtool/netlink.h
+++ b/net/ethtool/netlink.h
@@ -363,7 +363,7 @@ extern const struct nla_policy ethnl_features_set_policy[ETHTOOL_A_FEATURES_WANT
 extern const struct nla_policy ethnl_privflags_get_policy[ETHTOOL_A_PRIVFLAGS_HEADER + 1];
 extern const struct nla_policy ethnl_privflags_set_policy[ETHTOOL_A_PRIVFLAGS_FLAGS + 1];
 extern const struct nla_policy ethnl_rings_get_policy[ETHTOOL_A_RINGS_HEADER + 1];
-extern const struct nla_policy ethnl_rings_set_policy[ETHTOOL_A_RINGS_CQE_SIZE + 1];
+extern const struct nla_policy ethnl_rings_set_policy[ETHTOOL_A_RINGS_TX_PUSH + 1];
 extern const struct nla_policy ethnl_channels_get_policy[ETHTOOL_A_CHANNELS_HEADER + 1];
 extern const struct nla_policy ethnl_channels_set_policy[ETHTOOL_A_CHANNELS_COMBINED_COUNT + 1];
 extern const struct nla_policy ethnl_coalesce_get_policy[ETHTOOL_A_COALESCE_HEADER + 1];
diff --git a/net/ethtool/rings.c b/net/ethtool/rings.c
index 9f33c9689b5694..fa3ec8d438f7df 100644
--- a/net/ethtool/rings.c
+++ b/net/ethtool/rings.c
@@ -55,7 +55,8 @@ static int rings_reply_size(const struct ethnl_req_info *req_base,
 	       nla_total_size(sizeof(u32)) +	/* _RINGS_TX */
 	       nla_total_size(sizeof(u32)) +	/* _RINGS_RX_BUF_LEN */
 	       nla_total_size(sizeof(u8))  +	/* _RINGS_TCP_DATA_SPLIT */
-	       nla_total_size(sizeof(u32));	/* _RINGS_CQE_SIZE */
+	       nla_total_size(sizeof(u32)  +	/* _RINGS_CQE_SIZE */
+	       nla_total_size(sizeof(u8)));	/* _RINGS_TX_PUSH */
 }
 
 static int rings_fill_reply(struct sk_buff *skb,
@@ -94,7 +95,8 @@ static int rings_fill_reply(struct sk_buff *skb,
 	     (nla_put_u8(skb, ETHTOOL_A_RINGS_TCP_DATA_SPLIT,
 			 kr->tcp_data_split))) ||
 	    (kr->cqe_size &&
-	     (nla_put_u32(skb, ETHTOOL_A_RINGS_CQE_SIZE, kr->cqe_size))))
+	     (nla_put_u32(skb, ETHTOOL_A_RINGS_CQE_SIZE, kr->cqe_size))) ||
+	    nla_put_u8(skb, ETHTOOL_A_RINGS_TX_PUSH, !!kr->tx_push))
 		return -EMSGSIZE;
 
 	return 0;
@@ -123,6 +125,7 @@ const struct nla_policy ethnl_rings_set_policy[] = {
 	[ETHTOOL_A_RINGS_TX]			= { .type = NLA_U32 },
 	[ETHTOOL_A_RINGS_RX_BUF_LEN]            = NLA_POLICY_MIN(NLA_U32, 1),
 	[ETHTOOL_A_RINGS_CQE_SIZE]		= NLA_POLICY_MIN(NLA_U32, 1),
+	[ETHTOOL_A_RINGS_TX_PUSH]		= NLA_POLICY_MAX(NLA_U8, 1),
 };
 
 int ethnl_set_rings(struct sk_buff *skb, struct genl_info *info)
@@ -149,6 +152,33 @@ int ethnl_set_rings(struct sk_buff *skb, struct genl_info *info)
 	if (!ops->get_ringparam || !ops->set_ringparam)
 		goto out_dev;
 
+	if (tb[ETHTOOL_A_RINGS_RX_BUF_LEN] &&
+	    !(ops->supported_ring_params & ETHTOOL_RING_USE_RX_BUF_LEN)) {
+		ret = -EOPNOTSUPP;
+		NL_SET_ERR_MSG_ATTR(info->extack,
+				    tb[ETHTOOL_A_RINGS_RX_BUF_LEN],
+				    "setting rx buf len not supported");
+		goto out_dev;
+	}
+
+	if (tb[ETHTOOL_A_RINGS_CQE_SIZE] &&
+	    !(ops->supported_ring_params & ETHTOOL_RING_USE_CQE_SIZE)) {
+		ret = -EOPNOTSUPP;
+		NL_SET_ERR_MSG_ATTR(info->extack,
+				    tb[ETHTOOL_A_RINGS_CQE_SIZE],
+				    "setting cqe size not supported");
+		goto out_dev;
+	}
+
+	if (tb[ETHTOOL_A_RINGS_TX_PUSH] &&
+	    !(ops->supported_ring_params & ETHTOOL_RING_USE_TX_PUSH)) {
+		ret = -EOPNOTSUPP;
+		NL_SET_ERR_MSG_ATTR(info->extack,
+				    tb[ETHTOOL_A_RINGS_TX_PUSH],
+				    "setting tx push not supported");
+		goto out_dev;
+	}
+
 	rtnl_lock();
 	ret = ethnl_ops_begin(dev);
 	if (ret < 0)
@@ -165,6 +195,8 @@ int ethnl_set_rings(struct sk_buff *skb, struct genl_info *info)
 			 tb[ETHTOOL_A_RINGS_RX_BUF_LEN], &mod);
 	ethnl_update_u32(&kernel_ringparam.cqe_size,
 			 tb[ETHTOOL_A_RINGS_CQE_SIZE], &mod);
+	ethnl_update_u8(&kernel_ringparam.tx_push,
+			tb[ETHTOOL_A_RINGS_TX_PUSH], &mod);
 	ret = 0;
 	if (!mod)
 		goto out_ops;
@@ -187,24 +219,6 @@ int ethnl_set_rings(struct sk_buff *skb, struct genl_info *info)
 		goto out_ops;
 	}
 
-	if (kernel_ringparam.rx_buf_len != 0 &&
-	    !(ops->supported_ring_params & ETHTOOL_RING_USE_RX_BUF_LEN)) {
-		ret = -EOPNOTSUPP;
-		NL_SET_ERR_MSG_ATTR(info->extack,
-				    tb[ETHTOOL_A_RINGS_RX_BUF_LEN],
-				    "setting rx buf len not supported");
-		goto out_ops;
-	}
-
-	if (kernel_ringparam.cqe_size &&
-	    !(ops->supported_ring_params & ETHTOOL_RING_USE_CQE_SIZE)) {
-		ret = -EOPNOTSUPP;
-		NL_SET_ERR_MSG_ATTR(info->extack,
-				    tb[ETHTOOL_A_RINGS_CQE_SIZE],
-				    "setting cqe size not supported");
-		goto out_ops;
-	}
-
 	ret = dev->ethtool_ops->set_ringparam(dev, &ringparam,
 					      &kernel_ringparam, info->extack);
 	if (ret < 0)