Merge branch 'am65-cpsw-preemption-coalescing'

Roger Quadros says:

====================
net: ethernet: am65-cpsw: Add mqprio, frame preemption & coalescing

This series adds mqprio qdisc offload in channel mode,
Frame Preemption MAC merge support and RX/TX coalesing
for AM65 CPSW driver.

In v11 following changes were made
- Fix patch "net: ethernet: ti: am65-cpsw: add mqprio qdisc offload in channel mode"
by including units.h

Changelog information in each patch file.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>

Author: davem330

drivers/net/ethernet/ti/Kconfig

@@ -134,14 +134,16 @@ config TI_K3_AM65_CPTS
 	  protocol, Ethernet Enhanced Scheduled Traffic Operations (CPTS_ESTFn)
 	  and PCIe Subsystem Precision Time Measurement (PTM).
 
-config TI_AM65_CPSW_TAS
-	bool "Enable TAS offload in AM65 CPSW"
+config TI_AM65_CPSW_QOS
+	bool "Enable QoS offload features in AM65 CPSW"
 	depends on TI_K3_AM65_CPSW_NUSS && NET_SCH_TAPRIO && TI_K3_AM65_CPTS
 	help
-	  Say y here to support Time Aware Shaper(TAS) offload in AM65 CPSW.
-	  AM65 CPSW hardware supports Enhanced Scheduled Traffic (EST)
-	  defined in IEEE 802.1Q 2018. The EST scheduler runs on CPTS and the
-	  TAS/EST schedule is updated in the Fetch RAM memory of the CPSW.
+	  This option enables QoS offload features in AM65 CPSW like
+	  Time Aware Shaper (TAS) / Enhanced Scheduled Traffic (EST),
+	  MQPRIO qdisc offload and Frame-Preemption MAC Merge / Interspersing
+	  Express Traffic (IET).
+	  The EST scheduler runs on CPTS and the TAS/EST schedule is
+	  updated in the Fetch RAM memory of the CPSW.
 
 config TI_KEYSTONE_NETCP
 	tristate "TI Keystone NETCP Core Support"

drivers/net/ethernet/ti/Makefile

@@ -26,7 +26,8 @@ keystone_netcp_ethss-y := netcp_ethss.o netcp_sgmii.o netcp_xgbepcsr.o cpsw_ale.
 obj-$(CONFIG_TI_K3_CPPI_DESC_POOL) += k3-cppi-desc-pool.o
 
 obj-$(CONFIG_TI_K3_AM65_CPSW_NUSS) += ti-am65-cpsw-nuss.o
-ti-am65-cpsw-nuss-y := am65-cpsw-nuss.o cpsw_sl.o am65-cpsw-ethtool.o cpsw_ale.o am65-cpsw-qos.o
+ti-am65-cpsw-nuss-y := am65-cpsw-nuss.o cpsw_sl.o am65-cpsw-ethtool.o cpsw_ale.o
+ti-am65-cpsw-nuss-$(CONFIG_TI_AM65_CPSW_QOS) += am65-cpsw-qos.o
 ti-am65-cpsw-nuss-$(CONFIG_TI_K3_AM65_CPSW_SWITCHDEV) += am65-cpsw-switchdev.o
 obj-$(CONFIG_TI_K3_AM65_CPTS) += am65-cpts.o
 

drivers/net/ethernet/ti/am65-cpsw-ethtool.c

@@ -11,6 +11,7 @@
 #include <linux/pm_runtime.h>
 
 #include "am65-cpsw-nuss.h"
+#include "am65-cpsw-qos.h"
 #include "cpsw_ale.h"
 #include "am65-cpts.h"
 
@@ -670,6 +671,9 @@ static void am65_cpsw_get_eth_mac_stats(struct net_device *ndev,
 
 	stats = port->stat_base;
 
+	if (s->src != ETHTOOL_MAC_STATS_SRC_AGGREGATE)
+		return;
+
 	s->FramesTransmittedOK = readl_relaxed(&stats->tx_good_frames);
 	s->SingleCollisionFrames = readl_relaxed(&stats->tx_single_coll_frames);
 	s->MultipleCollisionFrames = readl_relaxed(&stats->tx_mult_coll_frames);
@@ -740,6 +744,240 @@ static int am65_cpsw_set_ethtool_priv_flags(struct net_device *ndev, u32 flags)
 	return 0;
 }
 
+static void am65_cpsw_port_iet_rx_enable(struct am65_cpsw_port *port, bool enable)
+{
+	u32 val;
+
+	val = readl(port->port_base + AM65_CPSW_PN_REG_CTL);
+	if (enable)
+		val |= AM65_CPSW_PN_CTL_IET_PORT_EN;
+	else
+		val &= ~AM65_CPSW_PN_CTL_IET_PORT_EN;
+
+	writel(val, port->port_base + AM65_CPSW_PN_REG_CTL);
+	am65_cpsw_iet_common_enable(port->common);
+}
+
+static void am65_cpsw_port_iet_tx_enable(struct am65_cpsw_port *port, bool enable)
+{
+	u32 val;
+
+	val = readl(port->port_base + AM65_CPSW_PN_REG_IET_CTRL);
+	if (enable)
+		val |= AM65_CPSW_PN_IET_MAC_PENABLE;
+	else
+		val &= ~AM65_CPSW_PN_IET_MAC_PENABLE;
+
+	writel(val, port->port_base + AM65_CPSW_PN_REG_IET_CTRL);
+}
+
+static int am65_cpsw_get_mm(struct net_device *ndev, struct ethtool_mm_state *state)
+{
+	struct am65_cpsw_port *port = am65_ndev_to_port(ndev);
+	struct am65_cpsw_ndev_priv *priv = netdev_priv(ndev);
+	u32 port_ctrl, iet_ctrl, iet_status;
+	u32 add_frag_size;
+
+	if (!IS_ENABLED(CONFIG_TI_AM65_CPSW_QOS))
+		return -EOPNOTSUPP;
+
+	mutex_lock(&priv->mm_lock);
+
+	iet_ctrl = readl(port->port_base + AM65_CPSW_PN_REG_IET_CTRL);
+	port_ctrl = readl(port->port_base + AM65_CPSW_PN_REG_CTL);
+
+	state->tx_enabled = !!(iet_ctrl & AM65_CPSW_PN_IET_MAC_PENABLE);
+	state->pmac_enabled = !!(port_ctrl & AM65_CPSW_PN_CTL_IET_PORT_EN);
+
+	iet_status = readl(port->port_base + AM65_CPSW_PN_REG_IET_STATUS);
+
+	if (iet_ctrl & AM65_CPSW_PN_IET_MAC_DISABLEVERIFY)
+		state->verify_status = ETHTOOL_MM_VERIFY_STATUS_DISABLED;
+	else if (iet_status & AM65_CPSW_PN_MAC_VERIFIED)
+		state->verify_status = ETHTOOL_MM_VERIFY_STATUS_SUCCEEDED;
+	else if (iet_status & AM65_CPSW_PN_MAC_VERIFY_FAIL)
+		state->verify_status = ETHTOOL_MM_VERIFY_STATUS_FAILED;
+	else
+		state->verify_status = ETHTOOL_MM_VERIFY_STATUS_UNKNOWN;
+
+	add_frag_size = AM65_CPSW_PN_IET_MAC_GET_ADDFRAGSIZE(iet_ctrl);
+	state->tx_min_frag_size = ethtool_mm_frag_size_add_to_min(add_frag_size);
+
+	/* Errata i2208: RX min fragment size cannot be less than 124 */
+	state->rx_min_frag_size = 124;
+
+	/* FPE active if common tx_enabled and verification success or disabled (forced) */
+	state->tx_active = state->tx_enabled &&
+			   (state->verify_status == ETHTOOL_MM_VERIFY_STATUS_SUCCEEDED ||
+			    state->verify_status == ETHTOOL_MM_VERIFY_STATUS_DISABLED);
+	state->verify_enabled = !(iet_ctrl & AM65_CPSW_PN_IET_MAC_DISABLEVERIFY);
+
+	state->verify_time = port->qos.iet.verify_time_ms;
+
+	/* 802.3-2018 clause 30.14.1.6, says that the aMACMergeVerifyTime
+	 * variable has a range between 1 and 128 ms inclusive. Limit to that.
+	 */
+	state->max_verify_time = 128;
+
+	mutex_unlock(&priv->mm_lock);
+
+	return 0;
+}
+
+static int am65_cpsw_set_mm(struct net_device *ndev, struct ethtool_mm_cfg *cfg,
+			    struct netlink_ext_ack *extack)
+{
+	struct am65_cpsw_port *port = am65_ndev_to_port(ndev);
+	struct am65_cpsw_ndev_priv *priv = netdev_priv(ndev);
+	struct am65_cpsw_iet *iet = &port->qos.iet;
+	u32 val, add_frag_size;
+	int err;
+
+	if (!IS_ENABLED(CONFIG_TI_AM65_CPSW_QOS))
+		return -EOPNOTSUPP;
+
+	err = ethtool_mm_frag_size_min_to_add(cfg->tx_min_frag_size, &add_frag_size, extack);
+	if (err)
+		return err;
+
+	mutex_lock(&priv->mm_lock);
+
+	if (cfg->pmac_enabled) {
+		/* change TX & RX FIFO MAX_BLKS as per TRM recommendation */
+		if (!iet->original_max_blks)
+			iet->original_max_blks = readl(port->port_base + AM65_CPSW_PN_REG_MAX_BLKS);
+
+		writel(AM65_CPSW_PN_TX_RX_MAX_BLKS_IET,
+		       port->port_base + AM65_CPSW_PN_REG_MAX_BLKS);
+	} else if (iet->original_max_blks) {
+		/* restore RX & TX FIFO MAX_BLKS */
+		writel(iet->original_max_blks,
+		       port->port_base + AM65_CPSW_PN_REG_MAX_BLKS);
+	}
+
+	am65_cpsw_port_iet_rx_enable(port, cfg->pmac_enabled);
+	am65_cpsw_port_iet_tx_enable(port, cfg->tx_enabled);
+
+	val = readl(port->port_base + AM65_CPSW_PN_REG_IET_CTRL);
+	if (cfg->verify_enabled) {
+		val &= ~AM65_CPSW_PN_IET_MAC_DISABLEVERIFY;
+		/* Reset Verify state machine. Verification won't start here.
+		 * Verification will be done once link-up.
+		 */
+		val |= AM65_CPSW_PN_IET_MAC_LINKFAIL;
+	} else {
+		val |= AM65_CPSW_PN_IET_MAC_DISABLEVERIFY;
+		/* Clear LINKFAIL to allow verify/response packets */
+		val &= ~AM65_CPSW_PN_IET_MAC_LINKFAIL;
+	}
+
+	val &= ~AM65_CPSW_PN_IET_MAC_MAC_ADDFRAGSIZE_MASK;
+	val |= AM65_CPSW_PN_IET_MAC_SET_ADDFRAGSIZE(add_frag_size);
+	writel(val, port->port_base + AM65_CPSW_PN_REG_IET_CTRL);
+
+	/* verify_timeout_count can only be set at valid link */
+	port->qos.iet.verify_time_ms = cfg->verify_time;
+
+	/* enable/disable preemption based on link status */
+	am65_cpsw_iet_commit_preemptible_tcs(port);
+
+	mutex_unlock(&priv->mm_lock);
+
+	return 0;
+}
+
+static void am65_cpsw_get_mm_stats(struct net_device *ndev,
+				   struct ethtool_mm_stats *s)
+{
+	struct am65_cpsw_port *port = am65_ndev_to_port(ndev);
+	void __iomem *base = port->stat_base;
+
+	s->MACMergeFrameAssOkCount = readl(base + AM65_CPSW_STATN_IET_RX_ASSEMBLY_OK);
+	s->MACMergeFrameAssErrorCount = readl(base + AM65_CPSW_STATN_IET_RX_ASSEMBLY_ERROR);
+	s->MACMergeFrameSmdErrorCount = readl(base + AM65_CPSW_STATN_IET_RX_SMD_ERROR);
+	/* CPSW Functional Spec states:
+	 * "The IET stat aMACMergeFragCountRx is derived by adding the
+	 *  Receive Assembly Error count to this value. i.e. AM65_CPSW_STATN_IET_RX_FRAG"
+	 */
+	s->MACMergeFragCountRx = readl(base + AM65_CPSW_STATN_IET_RX_FRAG) + s->MACMergeFrameAssErrorCount;
+	s->MACMergeFragCountTx = readl(base + AM65_CPSW_STATN_IET_TX_FRAG);
+	s->MACMergeHoldCount = readl(base + AM65_CPSW_STATN_IET_TX_HOLD);
+}
+
+static int am65_cpsw_get_coalesce(struct net_device *ndev, struct ethtool_coalesce *coal,
+				  struct kernel_ethtool_coalesce *kernel_coal,
+				  struct netlink_ext_ack *extack)
+{
+	struct am65_cpsw_common *common = am65_ndev_to_common(ndev);
+	struct am65_cpsw_tx_chn *tx_chn;
+
+	tx_chn = &common->tx_chns[0];
+
+	coal->rx_coalesce_usecs = common->rx_pace_timeout / 1000;
+	coal->tx_coalesce_usecs = tx_chn->tx_pace_timeout / 1000;
+
+	return 0;
+}
+
+static int am65_cpsw_get_per_queue_coalesce(struct net_device *ndev, u32 queue,
+					    struct ethtool_coalesce *coal)
+{
+	struct am65_cpsw_common *common = am65_ndev_to_common(ndev);
+	struct am65_cpsw_tx_chn *tx_chn;
+
+	if (queue >= AM65_CPSW_MAX_TX_QUEUES)
+		return -EINVAL;
+
+	tx_chn = &common->tx_chns[queue];
+
+	coal->tx_coalesce_usecs = tx_chn->tx_pace_timeout / 1000;
+
+	return 0;
+}
+
+static int am65_cpsw_set_coalesce(struct net_device *ndev, struct ethtool_coalesce *coal,
+				  struct kernel_ethtool_coalesce *kernel_coal,
+				  struct netlink_ext_ack *extack)
+{
+	struct am65_cpsw_common *common = am65_ndev_to_common(ndev);
+	struct am65_cpsw_tx_chn *tx_chn;
+
+	tx_chn = &common->tx_chns[0];
+
+	if (coal->rx_coalesce_usecs && coal->rx_coalesce_usecs < 20)
+		return -EINVAL;
+
+	if (coal->tx_coalesce_usecs && coal->tx_coalesce_usecs < 20)
+		return -EINVAL;
+
+	common->rx_pace_timeout = coal->rx_coalesce_usecs * 1000;
+	tx_chn->tx_pace_timeout = coal->tx_coalesce_usecs * 1000;
+
+	return 0;
+}
+
+static int am65_cpsw_set_per_queue_coalesce(struct net_device *ndev, u32 queue,
+					    struct ethtool_coalesce *coal)
+{
+	struct am65_cpsw_common *common = am65_ndev_to_common(ndev);
+	struct am65_cpsw_tx_chn *tx_chn;
+
+	if (queue >= AM65_CPSW_MAX_TX_QUEUES)
+		return -EINVAL;
+
+	tx_chn = &common->tx_chns[queue];
+
+	if (coal->tx_coalesce_usecs && coal->tx_coalesce_usecs < 20) {
+		dev_info(common->dev, "defaulting to min value of 20us for tx-usecs for tx-%u\n",
+			 queue);
+		coal->tx_coalesce_usecs = 20;
+	}
+
+	tx_chn->tx_pace_timeout = coal->tx_coalesce_usecs * 1000;
+
+	return 0;
+}
+
 const struct ethtool_ops am65_cpsw_ethtool_ops_slave = {
 	.begin			= am65_cpsw_ethtool_op_begin,
 	.complete		= am65_cpsw_ethtool_op_complete,
@@ -758,6 +996,11 @@ const struct ethtool_ops am65_cpsw_ethtool_ops_slave = {
 	.get_ts_info		= am65_cpsw_get_ethtool_ts_info,
 	.get_priv_flags		= am65_cpsw_get_ethtool_priv_flags,
 	.set_priv_flags		= am65_cpsw_set_ethtool_priv_flags,
+	.supported_coalesce_params = ETHTOOL_COALESCE_USECS,
+	.get_coalesce           = am65_cpsw_get_coalesce,
+	.set_coalesce           = am65_cpsw_set_coalesce,
+	.get_per_queue_coalesce = am65_cpsw_get_per_queue_coalesce,
+	.set_per_queue_coalesce = am65_cpsw_set_per_queue_coalesce,
 
 	.get_link		= ethtool_op_get_link,
 	.get_link_ksettings	= am65_cpsw_get_link_ksettings,
@@ -769,4 +1012,7 @@ const struct ethtool_ops am65_cpsw_ethtool_ops_slave = {
 	.get_eee		= am65_cpsw_get_eee,
 	.set_eee		= am65_cpsw_set_eee,
 	.nway_reset		= am65_cpsw_nway_reset,
+	.get_mm			= am65_cpsw_get_mm,
+	.set_mm			= am65_cpsw_set_mm,
+	.get_mm_stats		= am65_cpsw_get_mm_stats,
 };