[PATCH net-next 0/2] net: lan743x: speed-based and user-configurable interrupt moderation

[PATCH net-next 0/2] net: lan743x: speed-based and user-configurable interrupt moderation

[Dhanushkalyan G]

The lan743x driver programs a single fixed interrupt moderation timer
into all per-vector INT_MOD_CFG registers at interrupt open, regardless
of link speed and with no way for userspace to tune it.

This series makes the moderation timer follow the negotiated link speed
and then exposes it through ethtool:

Patch 1 sets the timer as a function of link speed in the MAC link-up
path (64 us at 2.5G, 150 us at 1G, 330 us at 100M/10M) and separates the
static vector-to-timer mapping (programmed once at interrupt open) from
the per-speed timer value.

Patch 2 adds ethtool .get_coalesce/.set_coalesce with
ETHTOOL_COALESCE_USECS and ETHTOOL_COALESCE_USE_ADAPTIVE_RX/TX. Adaptive
mode (the default) keeps the timer tracking the link speed; disabling it
pins the timer to a user supplied value. Because the hardware uses one
value for all vectors, RX and TX cannot be tuned independently, so
mismatched rx/tx requests are rejected.

Dhanushkalyan G (1):
  net: lan743x: add ethtool interrupt coalescing support

Thangaraj Samynathan (1):
  net: lan743x: set interrupt moderation timer based on link speed

 .../net/ethernet/microchip/lan743x_ethtool.c  | 70 ++++++++++++++++++
 drivers/net/ethernet/microchip/lan743x_main.c | 71 ++++++++++++++++---
 drivers/net/ethernet/microchip/lan743x_main.h | 14 +++-
 3 files changed, 144 insertions(+), 11 deletions(-)

-- 
2.34.1

[PATCH net-next 1/2] net: lan743x: set interrupt moderation timer based on link speed

[Dhanushkalyan G]

From: Thangaraj Samynathan <Thangaraj.S@microchip.com>

The driver programs a single fixed interrupt moderation timer into every
INT_MOD_CFG register at interrupt open, regardless of the negotiated link
speed. A fixed value is a poor trade-off across the supported speed
range: it is too coarse at low speeds and not aggressive enough at 2.5G.

Set the moderation timer as a function of the negotiated link speed in
the MAC link-up path: 64 us at 2.5G, 150 us at 1G, and 330 us at
100M/10M. The interrupt-vector-to-timer mapping (INT_MOD_MAP) is static,
so it stays programmed once at interrupt open; only the timer value
(INT_MOD_CFG) is updated, from the new lan743x_config_int_mod() helper,
when the link comes up.

Signed-off-by: Thangaraj Samynathan <Thangaraj.S@microchip.com>
Signed-off-by: Dhanushkalyan G <Dhanushkalyan.G@microchip.com>
---
 drivers/net/ethernet/microchip/lan743x_main.c | 51 ++++++++++++++-----
 drivers/net/ethernet/microchip/lan743x_main.h |  5 +-
 2 files changed, 42 insertions(+), 14 deletions(-)

diff --git a/drivers/net/ethernet/microchip/lan743x_main.c b/drivers/net/ethernet/microchip/lan743x_main.c
index f3332417162e..20d6a48ca0bb 100644
--- a/drivers/net/ethernet/microchip/lan743x_main.c
+++ b/drivers/net/ethernet/microchip/lan743x_main.c
@@ -627,18 +627,12 @@ static int lan743x_intr_open(struct lan743x_adapter *adapter)
 		lan743x_csr_write(adapter, INT_VEC_EN_SET,
 				  INT_VEC_EN_(0));
 
+	/* The interrupt-vector-to-moderation-timer mapping is static, so
+	 * program it once here. The timer values themselves are set later
+	 * (per link speed) via lan743x_config_int_mod().
+	 */
 	if (!(adapter->csr.flags & LAN743X_CSR_FLAG_IS_A0)) {
-		lan743x_csr_write(adapter, INT_MOD_CFG0, LAN743X_INT_MOD);
-		lan743x_csr_write(adapter, INT_MOD_CFG1, LAN743X_INT_MOD);
-		lan743x_csr_write(adapter, INT_MOD_CFG2, LAN743X_INT_MOD);
-		lan743x_csr_write(adapter, INT_MOD_CFG3, LAN743X_INT_MOD);
-		lan743x_csr_write(adapter, INT_MOD_CFG4, LAN743X_INT_MOD);
-		lan743x_csr_write(adapter, INT_MOD_CFG5, LAN743X_INT_MOD);
-		lan743x_csr_write(adapter, INT_MOD_CFG6, LAN743X_INT_MOD);
-		lan743x_csr_write(adapter, INT_MOD_CFG7, LAN743X_INT_MOD);
 		if (adapter->is_pci11x1x) {
-			lan743x_csr_write(adapter, INT_MOD_CFG8, LAN743X_INT_MOD);
-			lan743x_csr_write(adapter, INT_MOD_CFG9, LAN743X_INT_MOD);
 			lan743x_csr_write(adapter, INT_MOD_MAP0, 0x00007654);
 			lan743x_csr_write(adapter, INT_MOD_MAP1, 0x00003210);
 		} else {
@@ -3036,6 +3030,28 @@ static void lan743x_phylink_mac_link_down(struct phylink_config *config,
 	netif_tx_stop_all_queues(netdev);
 }
 
+/* Program the interrupt moderation timer value into the per-vector
+ * INT_MOD_CFG registers. Only the timer value is written here; the vector
+ * mapping (INT_MOD_MAP) is static and is set once at interrupt open.
+ */
+static void lan743x_config_int_mod(struct lan743x_adapter *adapter, u32 int_mod)
+{
+	if (!(adapter->csr.flags & LAN743X_CSR_FLAG_IS_A0)) {
+		lan743x_csr_write(adapter, INT_MOD_CFG0, int_mod);
+		lan743x_csr_write(adapter, INT_MOD_CFG1, int_mod);
+		lan743x_csr_write(adapter, INT_MOD_CFG2, int_mod);
+		lan743x_csr_write(adapter, INT_MOD_CFG3, int_mod);
+		lan743x_csr_write(adapter, INT_MOD_CFG4, int_mod);
+		lan743x_csr_write(adapter, INT_MOD_CFG5, int_mod);
+		lan743x_csr_write(adapter, INT_MOD_CFG6, int_mod);
+		lan743x_csr_write(adapter, INT_MOD_CFG7, int_mod);
+		if (adapter->is_pci11x1x) {
+			lan743x_csr_write(adapter, INT_MOD_CFG8, int_mod);
+			lan743x_csr_write(adapter, INT_MOD_CFG9, int_mod);
+		}
+	}
+}
+
 static void lan743x_phylink_mac_link_up(struct phylink_config *config,
 					struct phy_device *phydev,
 					unsigned int link_an_mode,
@@ -3045,6 +3061,7 @@ static void lan743x_phylink_mac_link_up(struct phylink_config *config,
 {
 	struct net_device *netdev = to_net_dev(config->dev);
 	struct lan743x_adapter *adapter = netdev_priv(netdev);
+	u32 int_mod;
 	int mac_cr;
 	u8 cap;
 
@@ -3053,12 +3070,18 @@ static void lan743x_phylink_mac_link_up(struct phylink_config *config,
 	 * Resulting value corresponds to SPEED_10
 	 */
 	mac_cr &= ~(MAC_CR_CFG_H_ | MAC_CR_CFG_L_);
-	if (speed == SPEED_2500)
+	if (speed == SPEED_2500) {
 		mac_cr |= MAC_CR_CFG_H_ | MAC_CR_CFG_L_;
-	else if (speed == SPEED_1000)
+		int_mod = LAN743X_INT_MOD_2_5G;
+	} else if (speed == SPEED_1000) {
 		mac_cr |= MAC_CR_CFG_H_;
-	else if (speed == SPEED_100)
+		int_mod = LAN743X_INT_MOD_1G;
+	} else if (speed == SPEED_100) {
 		mac_cr |= MAC_CR_CFG_L_;
+		int_mod = LAN743X_INT_MOD_100M;
+	} else {
+		int_mod = LAN743X_INT_MOD_10M;
+	}
 
 	if (duplex == DUPLEX_FULL)
 		mac_cr |= MAC_CR_DPX_;
@@ -3067,6 +3090,8 @@ static void lan743x_phylink_mac_link_up(struct phylink_config *config,
 
 	lan743x_csr_write(adapter, MAC_CR, mac_cr);
 
+	lan743x_config_int_mod(adapter, int_mod);
+
 	lan743x_ptp_update_latency(adapter, speed);
 
 	/* Flow Control operation */
diff --git a/drivers/net/ethernet/microchip/lan743x_main.h b/drivers/net/ethernet/microchip/lan743x_main.h
index 160d94a7cee6..c4a3de7fe107 100644
--- a/drivers/net/ethernet/microchip/lan743x_main.h
+++ b/drivers/net/ethernet/microchip/lan743x_main.h
@@ -861,7 +861,10 @@ struct lan743x_adapter;
 #define LAN743X_USED_RX_CHANNELS	(4)
 #define LAN743X_USED_TX_CHANNELS	(1)
 #define PCI11X1X_USED_TX_CHANNELS	(4)
-#define LAN743X_INT_MOD	(400)
+#define LAN743X_INT_MOD_2_5G		(64)
+#define LAN743X_INT_MOD_1G		(150)
+#define LAN743X_INT_MOD_100M		(330)
+#define LAN743X_INT_MOD_10M		(330)
 
 #if (LAN743X_USED_RX_CHANNELS > LAN743X_MAX_RX_CHANNELS)
 #error Invalid LAN743X_USED_RX_CHANNELS
-- 
2.34.1

[PATCH net-next 2/2] net: lan743x: add ethtool interrupt coalescing support

[Dhanushkalyan G]

Build on the speed-based interrupt moderation and let userspace inspect
and override it through ethtool's coalescing interface.

Add .get_coalesce and .set_coalesce and advertise ETHTOOL_COALESCE_USECS
together with ETHTOOL_COALESCE_USE_ADAPTIVE_RX/TX. The hardware programs
one moderation timer value across all per-vector INT_MOD_CFG registers,
so RX and TX cannot be tuned independently: get reports the same value
and adaptive state for both, and set rejects a request whose rx and tx
values (or adaptive flags) differ.

With adaptive coalescing enabled (the default) the timer keeps tracking
the link speed. Turning it off pins the timer to the user supplied value;
values are bounded by the register's maximum. Re-enabling adaptive mode
immediately restores the value for the current link speed. Writes to the
moderation registers are skipped when the value is unchanged.

Signed-off-by: Dhanushkalyan G <Dhanushkalyan.G@microchip.com>
---
 .../net/ethernet/microchip/lan743x_ethtool.c  | 70 +++++++++++++++++++
 drivers/net/ethernet/microchip/lan743x_main.c | 50 +++++++++----
 drivers/net/ethernet/microchip/lan743x_main.h |  9 +++
 3 files changed, 117 insertions(+), 12 deletions(-)

diff --git a/drivers/net/ethernet/microchip/lan743x_ethtool.c b/drivers/net/ethernet/microchip/lan743x_ethtool.c
index 9195419ecee0..76b60b3271be 100644
--- a/drivers/net/ethernet/microchip/lan743x_ethtool.c
+++ b/drivers/net/ethernet/microchip/lan743x_ethtool.c
@@ -1355,7 +1355,77 @@ static int lan743x_set_pauseparam(struct net_device *dev,
 	return phylink_ethtool_set_pauseparam(adapter->phylink, pause);
 }
 
+static int lan743x_ethtool_get_coalesce(struct net_device *netdev,
+					struct ethtool_coalesce *ec,
+					struct kernel_ethtool_coalesce *kernel_coal,
+					struct netlink_ext_ack *extack)
+{
+	struct lan743x_adapter *adapter = netdev_priv(netdev);
+
+	/* The driver programs the same value into every per-vector
+	 * INT_MOD_CFG register, so RX and TX moderation share one value;
+	 * report it for both.
+	 */
+	ec->rx_coalesce_usecs = adapter->int_mod;
+	ec->tx_coalesce_usecs = adapter->int_mod;
+	ec->use_adaptive_rx_coalesce = adapter->use_adaptive_mod;
+	ec->use_adaptive_tx_coalesce = adapter->use_adaptive_mod;
+
+	return 0;
+}
+
+static int lan743x_ethtool_set_coalesce(struct net_device *netdev,
+					struct ethtool_coalesce *ec,
+					struct kernel_ethtool_coalesce *kernel_coal,
+					struct netlink_ext_ack *extack)
+{
+	struct lan743x_adapter *adapter = netdev_priv(netdev);
+	u32 int_mod;
+
+	/* RX and TX share a single moderation value, so adaptive mode must be
+	 * enabled or disabled for both together.
+	 */
+	if (ec->use_adaptive_rx_coalesce != ec->use_adaptive_tx_coalesce) {
+		NL_SET_ERR_MSG_MOD(extack,
+				   "adaptive-rx and adaptive-tx must match");
+		return -EINVAL;
+	}
+
+	if (ec->use_adaptive_rx_coalesce) {
+		/* Resume speed-adaptive moderation and apply the value for the
+		 * current link speed right away.
+		 */
+		adapter->use_adaptive_mod = true;
+		int_mod = lan743x_get_int_mod(adapter->link_speed);
+	} else {
+		/* All per-vector INT_MOD_CFG registers get the same value, so
+		 * RX and TX moderation cannot differ; reject a mismatch.
+		 */
+		if (ec->rx_coalesce_usecs != ec->tx_coalesce_usecs) {
+			NL_SET_ERR_MSG_MOD(extack,
+					   "rx-usecs and tx-usecs must be equal");
+			return -EINVAL;
+		}
+		if (ec->rx_coalesce_usecs > LAN743X_INT_MOD_MAX) {
+			NL_SET_ERR_MSG_MOD(extack,
+					   "coalesce value exceeds maximum");
+			return -EINVAL;
+		}
+		adapter->use_adaptive_mod = false;
+		int_mod = ec->rx_coalesce_usecs;
+	}
+
+	lan743x_config_int_mod(adapter, int_mod);
+
+	return 0;
+}
+
 const struct ethtool_ops lan743x_ethtool_ops = {
+	.supported_coalesce_params = ETHTOOL_COALESCE_USECS |
+				     ETHTOOL_COALESCE_USE_ADAPTIVE_RX |
+				     ETHTOOL_COALESCE_USE_ADAPTIVE_TX,
+	.get_coalesce = lan743x_ethtool_get_coalesce,
+	.set_coalesce = lan743x_ethtool_set_coalesce,
 	.get_drvinfo = lan743x_ethtool_get_drvinfo,
 	.get_msglevel = lan743x_ethtool_get_msglevel,
 	.set_msglevel = lan743x_ethtool_set_msglevel,
diff --git a/drivers/net/ethernet/microchip/lan743x_main.c b/drivers/net/ethernet/microchip/lan743x_main.c
index 20d6a48ca0bb..9e1fb1e02346 100644
--- a/drivers/net/ethernet/microchip/lan743x_main.c
+++ b/drivers/net/ethernet/microchip/lan743x_main.c
@@ -3030,12 +3030,31 @@ static void lan743x_phylink_mac_link_down(struct phylink_config *config,
 	netif_tx_stop_all_queues(netdev);
 }
 
+/* Map a link speed to the interrupt moderation timer value to use. */
+u32 lan743x_get_int_mod(int speed)
+{
+	switch (speed) {
+	case SPEED_2500:
+		return LAN743X_INT_MOD_2_5G;
+	case SPEED_1000:
+		return LAN743X_INT_MOD_1G;
+	case SPEED_100:
+		return LAN743X_INT_MOD_100M;
+	default:
+		return LAN743X_INT_MOD_10M;
+	}
+}
+
 /* Program the interrupt moderation timer value into the per-vector
  * INT_MOD_CFG registers. Only the timer value is written here; the vector
  * mapping (INT_MOD_MAP) is static and is set once at interrupt open.
  */
-static void lan743x_config_int_mod(struct lan743x_adapter *adapter, u32 int_mod)
+void lan743x_config_int_mod(struct lan743x_adapter *adapter, u32 int_mod)
 {
+	/* Nothing to do if the value is already programmed in hardware. */
+	if (int_mod == adapter->int_mod)
+		return;
+
 	if (!(adapter->csr.flags & LAN743X_CSR_FLAG_IS_A0)) {
 		lan743x_csr_write(adapter, INT_MOD_CFG0, int_mod);
 		lan743x_csr_write(adapter, INT_MOD_CFG1, int_mod);
@@ -3049,6 +3068,7 @@ static void lan743x_config_int_mod(struct lan743x_adapter *adapter, u32 int_mod)
 			lan743x_csr_write(adapter, INT_MOD_CFG8, int_mod);
 			lan743x_csr_write(adapter, INT_MOD_CFG9, int_mod);
 		}
+		adapter->int_mod = int_mod;
 	}
 }
 
@@ -3061,7 +3081,6 @@ static void lan743x_phylink_mac_link_up(struct phylink_config *config,
 {
 	struct net_device *netdev = to_net_dev(config->dev);
 	struct lan743x_adapter *adapter = netdev_priv(netdev);
-	u32 int_mod;
 	int mac_cr;
 	u8 cap;
 
@@ -3070,18 +3089,12 @@ static void lan743x_phylink_mac_link_up(struct phylink_config *config,
 	 * Resulting value corresponds to SPEED_10
 	 */
 	mac_cr &= ~(MAC_CR_CFG_H_ | MAC_CR_CFG_L_);
-	if (speed == SPEED_2500) {
+	if (speed == SPEED_2500)
 		mac_cr |= MAC_CR_CFG_H_ | MAC_CR_CFG_L_;
-		int_mod = LAN743X_INT_MOD_2_5G;
-	} else if (speed == SPEED_1000) {
+	else if (speed == SPEED_1000)
 		mac_cr |= MAC_CR_CFG_H_;
-		int_mod = LAN743X_INT_MOD_1G;
-	} else if (speed == SPEED_100) {
+	else if (speed == SPEED_100)
 		mac_cr |= MAC_CR_CFG_L_;
-		int_mod = LAN743X_INT_MOD_100M;
-	} else {
-		int_mod = LAN743X_INT_MOD_10M;
-	}
 
 	if (duplex == DUPLEX_FULL)
 		mac_cr |= MAC_CR_DPX_;
@@ -3090,7 +3103,13 @@ static void lan743x_phylink_mac_link_up(struct phylink_config *config,
 
 	lan743x_csr_write(adapter, MAC_CR, mac_cr);
 
-	lan743x_config_int_mod(adapter, int_mod);
+	/* Track the link speed so a later ethtool request to re-enable
+	 * adaptive moderation can restore the speed-based value, and refresh
+	 * the timer now when adaptive moderation is active.
+	 */
+	adapter->link_speed = speed;
+	if (adapter->use_adaptive_mod)
+		lan743x_config_int_mod(adapter, lan743x_get_int_mod(speed));
 
 	lan743x_ptp_update_latency(adapter, speed);
 
@@ -3557,6 +3576,13 @@ static int lan743x_hardware_init(struct lan743x_adapter *adapter,
 	adapter->intr.irq = adapter->pdev->irq;
 	lan743x_csr_write(adapter, INT_EN_CLR, 0xFFFFFFFF);
 
+	/* Enable speed-adaptive interrupt moderation by default and program
+	 * the timer for the default (1G) value until the link comes up.
+	 */
+	adapter->use_adaptive_mod = true;
+	adapter->link_speed = SPEED_UNKNOWN;
+	lan743x_config_int_mod(adapter, LAN743X_INT_MOD_1G);
+
 	ret = lan743x_gpio_init(adapter);
 	if (ret)
 		return ret;
diff --git a/drivers/net/ethernet/microchip/lan743x_main.h b/drivers/net/ethernet/microchip/lan743x_main.h
index c4a3de7fe107..9e6ac8fdbf37 100644
--- a/drivers/net/ethernet/microchip/lan743x_main.h
+++ b/drivers/net/ethernet/microchip/lan743x_main.h
@@ -865,6 +865,7 @@ struct lan743x_adapter;
 #define LAN743X_INT_MOD_1G		(150)
 #define LAN743X_INT_MOD_100M		(330)
 #define LAN743X_INT_MOD_10M		(330)
+#define LAN743X_INT_MOD_MAX		(8191)
 
 #if (LAN743X_USED_RX_CHANNELS > LAN743X_MAX_RX_CHANNELS)
 #error Invalid LAN743X_USED_RX_CHANNELS
@@ -1091,6 +1092,12 @@ struct lan743x_adapter {
 	phy_interface_t		phy_interface;
 	struct phylink		*phylink;
 	struct phylink_config	phylink_config;
+	/* Interrupt moderation timer value currently programmed in hardware. */
+	u32			int_mod;
+	/* When set, the moderation timer tracks the link speed automatically. */
+	bool			use_adaptive_mod;
+	/* Last negotiated link speed, used to derive the adaptive value. */
+	int			link_speed;
 	int			rx_tstamp_filter;
 };
 
@@ -1212,5 +1219,7 @@ void lan743x_hs_syslock_release(struct lan743x_adapter *adapter);
 void lan743x_mac_flow_ctrl_set_enables(struct lan743x_adapter *adapter,
 				       bool tx_enable, bool rx_enable);
 int lan743x_sgmii_read(struct lan743x_adapter *adapter, u8 mmd, u16 addr);
+void lan743x_config_int_mod(struct lan743x_adapter *adapter, u32 int_mod);
+u32 lan743x_get_int_mod(int speed);
 
 #endif /* _LAN743X_H */
-- 
2.34.1

Re: [PATCH net-next 1/2] net: lan743x: set interrupt moderation timer based on link speed

[Jakub Kicinski]

On Mon, 3 Aug 2026 18:18:46 +0530 Dhanushkalyan G wrote:
> Set the moderation timer as a function of the negotiated link speed in
> the MAC link-up path: 64 us at 2.5G, 150 us at 1G, and 330 us at
> 100M/10M. The interrupt-vector-to-timer mapping (INT_MOD_MAP) is static,
> so it stays programmed once at interrupt open; only the timer value
> (INT_MOD_CFG) is updated, from the new lan743x_config_int_mod() helper,
> when the link comes up.

This patch should go away. With patch 2 user can chose the right
coalescing settings for themselves. adaptive coalescing means that 
the driver adjusts the coalescing settings based on _traffic_ not
link speed. Please don't (ab)use that for the trivial link speed based
heuristic. There's a library called DIM which implements the dynamic
algo which you can plug into your driver if you want (it even has docs)