[PATCH net-next 0/2] net: xilinx: axienet: Enable adaptive IRQ coalescing with DIM

[PATCH net-next 0/2] net: xilinx: axienet: Enable adaptive IRQ coalescing with DIM

[Sean Anderson]

To improve performance without sacrificing latency under low load,
enable DIM. While I appreciate not having to write the library myself, I
do think there are many unusual aspects to DIM, as detailed in the last
patch.

This series depends on [1].

[1] https://lore.kernel.org/netdev/20240903180059.4134461-1-sean.anderson@linux.dev/


Sean Anderson (2):
  net: xilinx: axienet: Support adjusting coalesce settings while
    running
  net: xilinx: axienet: Enable adaptive IRQ coalescing with DIM

 drivers/net/ethernet/xilinx/Kconfig           |   1 +
 drivers/net/ethernet/xilinx/xilinx_axienet.h  |  18 +-
 .../net/ethernet/xilinx/xilinx_axienet_main.c | 254 ++++++++++++++----
 3 files changed, 220 insertions(+), 53 deletions(-)

-- 
2.35.1.1320.gc452695387.dirty

[PATCH net-next 1/2] net: xilinx: axienet: Support adjusting coalesce settings while running

[Sean Anderson]

In preparation for adaptive IRQ coalescing, we first need to support
adjusting the settings at runtime. The existing code doesn't require any
locking because

- dma_start is the only function that modifies rx/tx_dma_cr. It is
  always called with IRQs and NAPI disabled, so nothing else is touching
  the hardware.
- The IRQs don't race with poll, since the latter is a softirq.
- The IRQs don't race with dma_stop since they both just clear the
  control registers.
- dma_stop doesn't race with poll since the former is called with NAPI
  disabled.

However, once we introduce another function that modifies rx/tx_dma_cr,
we need to have some locking to prevent races. Introduce two locks to
protect these variables and their registers.

The control register values are now generated where the coalescing
settings are set. This is because we need to hold the RTNL when reading
the settings, but axienet_dma_start cannot sleep.

Since axienet_dma_start writes the control settings when it runs, we don't
bother updating the CR registers when rx/tx_dma_started is false. This
prevents any issues from writing to the control registers in the middle
of a reset sequence.

Signed-off-by: Sean Anderson <sean.anderson@linux.dev>
---

 drivers/net/ethernet/xilinx/xilinx_axienet.h  |   8 +
 .../net/ethernet/xilinx/xilinx_axienet_main.c | 187 +++++++++++++-----
 2 files changed, 146 insertions(+), 49 deletions(-)

diff --git a/drivers/net/ethernet/xilinx/xilinx_axienet.h b/drivers/net/ethernet/xilinx/xilinx_axienet.h
index c301dd2ee083..66cb8aa5b716 100644
--- a/drivers/net/ethernet/xilinx/xilinx_axienet.h
+++ b/drivers/net/ethernet/xilinx/xilinx_axienet.h
@@ -484,7 +484,9 @@ struct skbuf_dma_descriptor {
  * @regs:	Base address for the axienet_local device address space
  * @dma_regs:	Base address for the axidma device address space
  * @napi_rx:	NAPI RX control structure
+ * @rx_cr_lock: Lock protecting @rx_dma_cr, its register, and @rx_dma_started
  * @rx_dma_cr:  Nominal content of RX DMA control register
+ * @rx_dma_started: Set when RX DMA is started
  * @rx_bd_v:	Virtual address of the RX buffer descriptor ring
  * @rx_bd_p:	Physical address(start address) of the RX buffer descr. ring
  * @rx_bd_num:	Size of RX buffer descriptor ring
@@ -494,7 +496,9 @@ struct skbuf_dma_descriptor {
  * @rx_bytes:	RX byte count for statistics
  * @rx_stat_sync: Synchronization object for RX stats
  * @napi_tx:	NAPI TX control structure
+ * @tx_cr_lock: Lock protecting @tx_dma_cr, its register, and @tx_dma_started
  * @tx_dma_cr:  Nominal content of TX DMA control register
+ * @tx_dma_started: Set when TX DMA is started
  * @tx_bd_v:	Virtual address of the TX buffer descriptor ring
  * @tx_bd_p:	Physical address(start address) of the TX buffer descr. ring
  * @tx_bd_num:	Size of TX buffer descriptor ring
@@ -566,7 +570,9 @@ struct axienet_local {
 	void __iomem *dma_regs;
 
 	struct napi_struct napi_rx;
+	spinlock_t rx_cr_lock;
 	u32 rx_dma_cr;
+	bool rx_dma_started;
 	struct axidma_bd *rx_bd_v;
 	dma_addr_t rx_bd_p;
 	u32 rx_bd_num;
@@ -576,7 +582,9 @@ struct axienet_local {
 	struct u64_stats_sync rx_stat_sync;
 
 	struct napi_struct napi_tx;
+	spinlock_t tx_cr_lock;
 	u32 tx_dma_cr;
+	bool tx_dma_started;
 	struct axidma_bd *tx_bd_v;
 	dma_addr_t tx_bd_p;
 	u32 tx_bd_num;
diff --git a/drivers/net/ethernet/xilinx/xilinx_axienet_main.c b/drivers/net/ethernet/xilinx/xilinx_axienet_main.c
index fd5cca4f5499..7bd109b77afc 100644
--- a/drivers/net/ethernet/xilinx/xilinx_axienet_main.c
+++ b/drivers/net/ethernet/xilinx/xilinx_axienet_main.c
@@ -224,25 +224,41 @@ static void axienet_dma_bd_release(struct net_device *ndev)
 }
 
 /**
- * axienet_usec_to_timer - Calculate IRQ delay timer value
- * @lp:		Pointer to the axienet_local structure
- * @coalesce_usec: Microseconds to convert into timer value
+ * axienet_calc_cr() - Calculate control register value
+ * @lp: Device private data
+ * @coalesce_count: Number of completions before an interrupt
+ * @coalesce_usec: Microseconds after the last completion before an interrupt
+ *
+ * Calculate a control register value based on the coalescing settings. The
+ * run/stop bit is not set.
  */
-static u32 axienet_usec_to_timer(struct axienet_local *lp, u32 coalesce_usec)
+static u32 axienet_calc_cr(struct axienet_local *lp, u32 coalesce_count, u32 coalesce_usec)
 {
-	u32 result;
-	u64 clk_rate = 125000000; /* arbitrary guess if no clock rate set */
+	u32 cr;
 
-	if (lp->axi_clk)
-		clk_rate = clk_get_rate(lp->axi_clk);
+	coalesce_count = min(coalesce_count, 255);
+	cr = (coalesce_count << XAXIDMA_COALESCE_SHIFT) | XAXIDMA_IRQ_IOC_MASK |
+	     XAXIDMA_IRQ_ERROR_MASK;
+	/* Only set interrupt delay timer if not generating an interrupt on
+	 * the first packet. Otherwise leave at 0 to disable delay interrupt.
+	 */
+	if (coalesce_count > 1) {
+		u64 clk_rate = 125000000; /* arbitrary guess if no clock rate set */
+		u32 timer;
 
-	/* 1 Timeout Interval = 125 * (clock period of SG clock) */
-	result = DIV64_U64_ROUND_CLOSEST((u64)coalesce_usec * clk_rate,
-					 (u64)125000000);
-	if (result > 255)
-		result = 255;
+		if (lp->axi_clk)
+			clk_rate = clk_get_rate(lp->axi_clk);
 
-	return result;
+		/* 1 Timeout Interval = 125 * (clock period of SG clock) */
+		timer = DIV64_U64_ROUND_CLOSEST((u64)coalesce_usec * clk_rate,
+						(u64)125000000);
+		if (timer > 255)
+			timer = 255;
+
+		cr |= (timer << XAXIDMA_DELAY_SHIFT) | XAXIDMA_IRQ_DELAY_MASK;
+	}
+
+	return cr;
 }
 
 /**
@@ -251,32 +267,12 @@ static u32 axienet_usec_to_timer(struct axienet_local *lp, u32 coalesce_usec)
  */
 static void axienet_dma_start(struct axienet_local *lp)
 {
+	spin_lock_irq(&lp->rx_cr_lock);
+
 	/* Start updating the Rx channel control register */
-	lp->rx_dma_cr = (min(lp->coalesce_count_rx, 255) <<
-			 XAXIDMA_COALESCE_SHIFT) |
-			XAXIDMA_IRQ_IOC_MASK | XAXIDMA_IRQ_ERROR_MASK;
-	/* Only set interrupt delay timer if not generating an interrupt on
-	 * the first RX packet. Otherwise leave at 0 to disable delay interrupt.
-	 */
-	if (lp->coalesce_count_rx > 1)
-		lp->rx_dma_cr |= (axienet_usec_to_timer(lp, lp->coalesce_usec_rx)
-					<< XAXIDMA_DELAY_SHIFT) |
-				 XAXIDMA_IRQ_DELAY_MASK;
+	lp->rx_dma_cr &= ~XAXIDMA_CR_RUNSTOP_MASK;
 	axienet_dma_out32(lp, XAXIDMA_RX_CR_OFFSET, lp->rx_dma_cr);
 
-	/* Start updating the Tx channel control register */
-	lp->tx_dma_cr = (min(lp->coalesce_count_tx, 255) <<
-			 XAXIDMA_COALESCE_SHIFT) |
-			XAXIDMA_IRQ_IOC_MASK | XAXIDMA_IRQ_ERROR_MASK;
-	/* Only set interrupt delay timer if not generating an interrupt on
-	 * the first TX packet. Otherwise leave at 0 to disable delay interrupt.
-	 */
-	if (lp->coalesce_count_tx > 1)
-		lp->tx_dma_cr |= (axienet_usec_to_timer(lp, lp->coalesce_usec_tx)
-					<< XAXIDMA_DELAY_SHIFT) |
-				 XAXIDMA_IRQ_DELAY_MASK;
-	axienet_dma_out32(lp, XAXIDMA_TX_CR_OFFSET, lp->tx_dma_cr);
-
 	/* Populate the tail pointer and bring the Rx Axi DMA engine out of
 	 * halted state. This will make the Rx side ready for reception.
 	 */
@@ -285,6 +281,14 @@ static void axienet_dma_start(struct axienet_local *lp)
 	axienet_dma_out32(lp, XAXIDMA_RX_CR_OFFSET, lp->rx_dma_cr);
 	axienet_dma_out_addr(lp, XAXIDMA_RX_TDESC_OFFSET, lp->rx_bd_p +
 			     (sizeof(*lp->rx_bd_v) * (lp->rx_bd_num - 1)));
+	lp->rx_dma_started = true;
+
+	spin_unlock_irq(&lp->rx_cr_lock);
+	spin_lock_irq(&lp->tx_cr_lock);
+
+	/* Start updating the Tx channel control register */
+	lp->tx_dma_cr &= ~XAXIDMA_CR_RUNSTOP_MASK;
+	axienet_dma_out32(lp, XAXIDMA_TX_CR_OFFSET, lp->tx_dma_cr);
 
 	/* Write to the RS (Run-stop) bit in the Tx channel control register.
 	 * Tx channel is now ready to run. But only after we write to the
@@ -293,6 +297,9 @@ static void axienet_dma_start(struct axienet_local *lp)
 	axienet_dma_out_addr(lp, XAXIDMA_TX_CDESC_OFFSET, lp->tx_bd_p);
 	lp->tx_dma_cr |= XAXIDMA_CR_RUNSTOP_MASK;
 	axienet_dma_out32(lp, XAXIDMA_TX_CR_OFFSET, lp->tx_dma_cr);
+	lp->tx_dma_started = true;
+
+	spin_unlock_irq(&lp->tx_cr_lock);
 }
 
 /**
@@ -628,14 +635,22 @@ static void axienet_dma_stop(struct axienet_local *lp)
 	int count;
 	u32 cr, sr;
 
-	cr = axienet_dma_in32(lp, XAXIDMA_RX_CR_OFFSET);
-	cr &= ~(XAXIDMA_CR_RUNSTOP_MASK | XAXIDMA_IRQ_ALL_MASK);
+	spin_lock_irq(&lp->rx_cr_lock);
+
+	cr = lp->rx_dma_cr & ~(XAXIDMA_CR_RUNSTOP_MASK | XAXIDMA_IRQ_ALL_MASK);
 	axienet_dma_out32(lp, XAXIDMA_RX_CR_OFFSET, cr);
+	lp->rx_dma_started = false;
+
+	spin_unlock_irq(&lp->rx_cr_lock);
 	synchronize_irq(lp->rx_irq);
 
-	cr = axienet_dma_in32(lp, XAXIDMA_TX_CR_OFFSET);
-	cr &= ~(XAXIDMA_CR_RUNSTOP_MASK | XAXIDMA_IRQ_ALL_MASK);
+	spin_lock_irq(&lp->tx_cr_lock);
+
+	cr = lp->tx_dma_cr & ~(XAXIDMA_CR_RUNSTOP_MASK | XAXIDMA_IRQ_ALL_MASK);
 	axienet_dma_out32(lp, XAXIDMA_TX_CR_OFFSET, cr);
+	lp->tx_dma_started = false;
+
+	spin_unlock_irq(&lp->tx_cr_lock);
 	synchronize_irq(lp->tx_irq);
 
 	/* Give DMAs a chance to halt gracefully */
@@ -975,11 +990,15 @@ static int axienet_tx_poll(struct napi_struct *napi, int budget)
 	}
 
 	if (packets < budget && napi_complete_done(napi, packets)) {
+		unsigned long flags;
+
 		/* Re-enable TX completion interrupts. This should
 		 * cause an immediate interrupt if any TX packets are
 		 * already pending.
 		 */
+		spin_lock_irqsave(&lp->tx_cr_lock, flags);
 		axienet_dma_out32(lp, XAXIDMA_TX_CR_OFFSET, lp->tx_dma_cr);
+		spin_unlock_irqrestore(&lp->tx_cr_lock, flags);
 	}
 	return packets;
 }
@@ -1241,11 +1260,15 @@ static int axienet_rx_poll(struct napi_struct *napi, int budget)
 		axienet_dma_out_addr(lp, XAXIDMA_RX_TDESC_OFFSET, tail_p);
 
 	if (packets < budget && napi_complete_done(napi, packets)) {
+		unsigned long flags;
+
 		/* Re-enable RX completion interrupts. This should
 		 * cause an immediate interrupt if any RX packets are
 		 * already pending.
 		 */
+		spin_lock_irqsave(&lp->rx_cr_lock, flags);
 		axienet_dma_out32(lp, XAXIDMA_RX_CR_OFFSET, lp->rx_dma_cr);
+		spin_unlock_irqrestore(&lp->rx_cr_lock, flags);
 	}
 	return packets;
 }
@@ -1283,10 +1306,14 @@ static irqreturn_t axienet_tx_irq(int irq, void *_ndev)
 		/* Disable further TX completion interrupts and schedule
 		 * NAPI to handle the completions.
 		 */
-		u32 cr = lp->tx_dma_cr;
+		unsigned long flags;
+		u32 cr;
 
+		spin_lock_irqsave(&lp->tx_cr_lock, flags);
+		cr = lp->tx_dma_cr;
 		cr &= ~(XAXIDMA_IRQ_IOC_MASK | XAXIDMA_IRQ_DELAY_MASK);
 		axienet_dma_out32(lp, XAXIDMA_TX_CR_OFFSET, cr);
+		spin_unlock_irqrestore(&lp->tx_cr_lock, flags);
 
 		napi_schedule(&lp->napi_tx);
 	}
@@ -1327,10 +1354,14 @@ static irqreturn_t axienet_rx_irq(int irq, void *_ndev)
 		/* Disable further RX completion interrupts and schedule
 		 * NAPI receive.
 		 */
-		u32 cr = lp->rx_dma_cr;
+		u32 cr;
+		unsigned long flags;
 
+		spin_lock_irqsave(&lp->rx_cr_lock, flags);
+		cr = lp->rx_dma_cr;
 		cr &= ~(XAXIDMA_IRQ_IOC_MASK | XAXIDMA_IRQ_DELAY_MASK);
 		axienet_dma_out32(lp, XAXIDMA_RX_CR_OFFSET, cr);
+		spin_unlock_irqrestore(&lp->rx_cr_lock, flags);
 
 		napi_schedule(&lp->napi_rx);
 	}
@@ -1992,6 +2023,62 @@ axienet_ethtools_set_pauseparam(struct net_device *ndev,
 	return phylink_ethtool_set_pauseparam(lp->phylink, epauseparm);
 }
 
+/**
+ * axienet_update_coalesce_rx() - Update RX coalesce settings
+ * @lp: Device private data
+ */
+static void axienet_update_coalesce_rx(struct axienet_local *lp)
+{
+	u32 cr = axienet_calc_cr(lp, lp->coalesce_count_rx,
+				 lp->coalesce_usec_rx);
+
+	spin_lock_irq(&lp->rx_cr_lock);
+	lp->rx_dma_cr &= XAXIDMA_CR_RUNSTOP_MASK;
+	lp->rx_dma_cr |= cr;
+	/* If DMA isn't started, then the settings will be applied the next
+	 * time dma_start() is called.
+	 */
+	if (lp->rx_dma_started) {
+		u32 reg = axienet_dma_in32(lp, XAXIDMA_RX_CR_OFFSET);
+
+		/* Don't enable IRQs if they are disabled by NAPI */
+		if (reg & XAXIDMA_IRQ_ALL_MASK)
+			cr = lp->rx_dma_cr;
+		else
+			cr = lp->rx_dma_cr & ~XAXIDMA_IRQ_ALL_MASK;
+		axienet_dma_out32(lp, XAXIDMA_RX_CR_OFFSET, cr);
+	}
+	spin_unlock_irq(&lp->rx_cr_lock);
+}
+
+/**
+ * axienet_update_coalesce_tx() - Update TX coalesce settings
+ * @lp: Device private data
+ */
+static void axienet_update_coalesce_tx(struct axienet_local *lp)
+{
+	u32 cr = axienet_calc_cr(lp, lp->coalesce_count_tx,
+				 lp->coalesce_usec_tx);
+
+	spin_lock_irq(&lp->tx_cr_lock);
+	lp->tx_dma_cr &= XAXIDMA_CR_RUNSTOP_MASK;
+	lp->tx_dma_cr |= cr;
+	/* If DMA isn't started, then the settings will be applied the next
+	 * time dma_start() is called.
+	 */
+	if (lp->tx_dma_started) {
+		u32 reg = axienet_dma_in32(lp, XAXIDMA_RX_CR_OFFSET);
+
+		/* Don't enable IRQs if they are disabled by NAPI */
+		if (reg & XAXIDMA_IRQ_ALL_MASK)
+			cr = lp->tx_dma_cr;
+		else
+			cr = lp->tx_dma_cr & ~XAXIDMA_IRQ_ALL_MASK;
+		axienet_dma_out32(lp, XAXIDMA_TX_CR_OFFSET, cr);
+	}
+	spin_unlock_irq(&lp->tx_cr_lock);
+}
+
 /**
  * axienet_ethtools_get_coalesce - Get DMA interrupt coalescing count.
  * @ndev:	Pointer to net_device structure
@@ -2041,12 +2128,6 @@ axienet_ethtools_set_coalesce(struct net_device *ndev,
 {
 	struct axienet_local *lp = netdev_priv(ndev);
 
-	if (netif_running(ndev)) {
-		NL_SET_ERR_MSG(extack,
-			       "Please stop netif before applying configuration");
-		return -EBUSY;
-	}
-
 	if (ecoalesce->rx_max_coalesced_frames)
 		lp->coalesce_count_rx = ecoalesce->rx_max_coalesced_frames;
 	if (ecoalesce->rx_coalesce_usecs)
@@ -2056,6 +2137,8 @@ axienet_ethtools_set_coalesce(struct net_device *ndev,
 	if (ecoalesce->tx_coalesce_usecs)
 		lp->coalesce_usec_tx = ecoalesce->tx_coalesce_usecs;
 
+	axienet_update_coalesce_rx(lp);
+	axienet_update_coalesce_tx(lp);
 	return 0;
 }
 
@@ -2840,10 +2923,16 @@ static int axienet_probe(struct platform_device *pdev)
 		axienet_set_mac_address(ndev, NULL);
 	}
 
+	spin_lock_init(&lp->rx_cr_lock);
+	spin_lock_init(&lp->tx_cr_lock);
 	lp->coalesce_count_rx = XAXIDMA_DFT_RX_THRESHOLD;
 	lp->coalesce_count_tx = XAXIDMA_DFT_TX_THRESHOLD;
 	lp->coalesce_usec_rx = XAXIDMA_DFT_RX_USEC;
 	lp->coalesce_usec_tx = XAXIDMA_DFT_TX_USEC;
+	lp->rx_dma_cr = axienet_calc_cr(lp, lp->coalesce_count_rx,
+					lp->coalesce_usec_rx);
+	lp->tx_dma_cr = axienet_calc_cr(lp, lp->coalesce_count_tx,
+					lp->coalesce_usec_tx);
 
 	ret = axienet_mdio_setup(lp);
 	if (ret)
-- 
2.35.1.1320.gc452695387.dirty

[PATCH net-next 2/2] net: xilinx: axienet: Enable adaptive IRQ coalescing with DIM

[Sean Anderson]

The default RX IRQ coalescing settings of one IRQ per packet can represent
a significant CPU load. However, increasing the coalescing unilaterally
can result in undesirable latency under low load. Adaptive IRQ
coalescing with DIM offers a way to adjust the coalescing settings based
on load.

This device only supports "CQE" mode [1], where each packet resets the
timer. Therefore, an interrupt is fired either when we receive
coalesce_count_rx packets or when the interface is idle for
coalesce_usec_rx. With this in mind, consider the following scenarios:

Link saturated
    Here we want to set coalesce_count_rx to a large value, in order to
    coalesce more packets and reduce CPU load. coalesce_usec_rx should
    be set to at least the time for one packet. Otherwise the link will
    be "idle" and we will get an interrupt for each packet anyway.

Bursts of packets
    Each burst should be coalesced into a single interrupt, although it
    may be prudent to reduce coalesce_count_rx for better latency.
    coalesce_usec_rx should be set to at least the time for one packet
    so bursts are coalesced. However, additional time beyond the packet
    time will just increase latency at the end of a burst.

Sporadic packets
    Due to low load, we can set coalesce_count_rx to 1 in order to
    reduce latency to the minimum. coalesce_usec_rx does not matter in
    this case.

Based on this analysis, I expected the CQE profiles to look something
like

	usec =  0, pkts = 1   // Low load
	usec = 16, pkts = 4
	usec = 16, pkts = 16
	usec = 16, pkts = 64
	usec = 16, pkts = 256 // High load

Where usec is set to 16 to be a few us greater than the 12.3 us packet
time of a 1500 MTU packet at 1 GBit/s. However, the CQE profile is
instead

	usec =  2, pkts = 256 // Low load
	usec =  8, pkts = 128
	usec = 16, pkts =  64
	usec = 32, pkts =  64
	usec = 64, pkts =  64 // High load

I found this very surprising. The number of coalesced packets
*decreases* as load increases. But as load increases we have more
opportunities to coalesce packets without affecting latency as much.
Additionally, the profile *increases* the usec as the load increases.
But as load increases, the gaps between packets will tend to become
smaller, making it possible to *decrease* usec for better latency at the
end of a "burst".

I consider the default CQE profile unsuitable for this NIC. Therefore,
we use the first profile outlined in this commit instead.
coalesce_usec_rx is set to 16 by default, but the user can customize it.
This may be necessary if they are using jumbo frames. I think adjusting
the profile times based on the link speed/mtu would be good improvement
for generic DIM.

In addition to the above profile problems, I noticed the following
additional issues with DIM while testing:

- DIM tends to "wander" when at low load, since the performance gradient
  is pretty flat. If you only have 10p/ms anyway then adjusting the
  coalescing settings will not affect throughput very much.
- DIM takes a long time to adjust back to low indices when load is
  decreased following a period of high load. This is because it only
  re-evaluates its settings once every 64 interrupts. However, at low
  load 64 interrupts can be several seconds.

Finally: performance. This patch increases receive throughput with
iperf3 from 840 Mbits/sec to 938 Mbits/sec, decreases interrupts from
69920/sec to 316/sec, and decreases CPU utilization (4x Cortex-A53) from
43% to 9%. I did not notice an increase in latency with this patch
applied.

[1] Who names this stuff?

Signed-off-by: Sean Anderson <sean.anderson@linux.dev>
---
Heng, maybe you have some comments on DIM regarding the above?

 drivers/net/ethernet/xilinx/Kconfig           |  1 +
 drivers/net/ethernet/xilinx/xilinx_axienet.h  | 10 ++-
 .../net/ethernet/xilinx/xilinx_axienet_main.c | 71 +++++++++++++++++--
 3 files changed, 76 insertions(+), 6 deletions(-)

diff --git a/drivers/net/ethernet/xilinx/Kconfig b/drivers/net/ethernet/xilinx/Kconfig
index 35d96c633a33..7502214cc7d5 100644
--- a/drivers/net/ethernet/xilinx/Kconfig
+++ b/drivers/net/ethernet/xilinx/Kconfig
@@ -28,6 +28,7 @@ config XILINX_AXI_EMAC
 	depends on HAS_IOMEM
 	depends on XILINX_DMA
 	select PHYLINK
+	select DIMLIB
 	help
 	  This driver supports the 10/100/1000 Ethernet from Xilinx for the
 	  AXI bus interface used in Xilinx Virtex FPGAs and Soc's.
diff --git a/drivers/net/ethernet/xilinx/xilinx_axienet.h b/drivers/net/ethernet/xilinx/xilinx_axienet.h
index 66cb8aa5b716..68303ece8285 100644
--- a/drivers/net/ethernet/xilinx/xilinx_axienet.h
+++ b/drivers/net/ethernet/xilinx/xilinx_axienet.h
@@ -9,6 +9,7 @@
 #ifndef XILINX_AXIENET_H
 #define XILINX_AXIENET_H
 
+#include <linux/dim.h>
 #include <linux/netdevice.h>
 #include <linux/spinlock.h>
 #include <linux/interrupt.h>
@@ -123,8 +124,7 @@
 /* Default TX/RX Threshold and delay timer values for SGDMA mode */
 #define XAXIDMA_DFT_TX_THRESHOLD	24
 #define XAXIDMA_DFT_TX_USEC		50
-#define XAXIDMA_DFT_RX_THRESHOLD	1
-#define XAXIDMA_DFT_RX_USEC		50
+#define XAXIDMA_DFT_RX_USEC		16
 
 #define XAXIDMA_BD_CTRL_TXSOF_MASK	0x08000000 /* First tx packet */
 #define XAXIDMA_BD_CTRL_TXEOF_MASK	0x04000000 /* Last tx packet */
@@ -484,6 +484,9 @@ struct skbuf_dma_descriptor {
  * @regs:	Base address for the axienet_local device address space
  * @dma_regs:	Base address for the axidma device address space
  * @napi_rx:	NAPI RX control structure
+ * @rx_dim:     DIM state for the receive queue
+ * @rx_irqs:    Number of interrupts
+ * @rx_dim_enabled: Whether DIM is enabled or not
  * @rx_cr_lock: Lock protecting @rx_dma_cr, its register, and @rx_dma_started
  * @rx_dma_cr:  Nominal content of RX DMA control register
  * @rx_dma_started: Set when RX DMA is started
@@ -570,6 +573,9 @@ struct axienet_local {
 	void __iomem *dma_regs;
 
 	struct napi_struct napi_rx;
+	struct dim rx_dim;
+	bool rx_dim_enabled;
+	u16 rx_irqs;
 	spinlock_t rx_cr_lock;
 	u32 rx_dma_cr;
 	bool rx_dma_started;
diff --git a/drivers/net/ethernet/xilinx/xilinx_axienet_main.c b/drivers/net/ethernet/xilinx/xilinx_axienet_main.c
index 7bd109b77afc..24d434f99bce 100644
--- a/drivers/net/ethernet/xilinx/xilinx_axienet_main.c
+++ b/drivers/net/ethernet/xilinx/xilinx_axienet_main.c
@@ -1262,6 +1262,18 @@ static int axienet_rx_poll(struct napi_struct *napi, int budget)
 	if (packets < budget && napi_complete_done(napi, packets)) {
 		unsigned long flags;
 
+		if (READ_ONCE(lp->rx_dim_enabled)) {
+			struct dim_sample sample = {
+				.time = ktime_get(),
+				/* Safe because we are the only writer */
+				.pkt_ctr = u64_stats_read(&lp->rx_packets),
+				.byte_ctr = u64_stats_read(&lp->rx_bytes),
+				.event_ctr = lp->rx_irqs,
+			};
+
+			net_dim(&lp->rx_dim, sample);
+		}
+
 		/* Re-enable RX completion interrupts. This should
 		 * cause an immediate interrupt if any RX packets are
 		 * already pending.
@@ -1364,6 +1376,7 @@ static irqreturn_t axienet_rx_irq(int irq, void *_ndev)
 		spin_unlock_irqrestore(&lp->rx_cr_lock, flags);
 
 		napi_schedule(&lp->napi_rx);
+		lp->rx_irqs++;
 	}
 
 	return IRQ_HANDLED;
@@ -1588,6 +1601,7 @@ static int axienet_init_legacy_dma(struct net_device *ndev)
 	napi_disable(&lp->napi_tx);
 	napi_disable(&lp->napi_rx);
 	cancel_work_sync(&lp->dma_err_task);
+	cancel_work_sync(&lp->rx_dim.work);
 	dev_err(lp->dev, "request_irq() failed\n");
 	return ret;
 }
@@ -1679,6 +1693,7 @@ static int axienet_stop(struct net_device *ndev)
 		napi_disable(&lp->napi_rx);
 	}
 
+	cancel_work_sync(&lp->rx_dim.work);
 	cancel_delayed_work_sync(&lp->stats_work);
 
 	phylink_stop(lp->phylink);
@@ -2051,6 +2066,32 @@ static void axienet_update_coalesce_rx(struct axienet_local *lp)
 	spin_unlock_irq(&lp->rx_cr_lock);
 }
 
+/**
+ * axienet_dim_coalesce_rx() - Update RX coalesce settings from DIM
+ * @lp: Device private data
+ */
+static void axienet_dim_coalesce_rx(struct axienet_local *lp)
+{
+	lp->coalesce_count_rx = 1 << (lp->rx_dim.profile_ix << 1);
+}
+
+/**
+ * axienet_rx_dim_work() - Adjust RX DIM settings
+ * @work: The work struct
+ */
+static void axienet_rx_dim_work(struct work_struct *work)
+{
+	struct axienet_local *lp =
+		container_of(work, struct axienet_local, rx_dim.work);
+
+	rtnl_lock();
+	axienet_dim_coalesce_rx(lp);
+	axienet_update_coalesce_rx(lp);
+	rtnl_unlock();
+
+	lp->rx_dim.state = DIM_START_MEASURE;
+}
+
 /**
  * axienet_update_coalesce_tx() - Update TX coalesce settings
  * @lp: Device private data
@@ -2100,6 +2141,7 @@ axienet_ethtools_get_coalesce(struct net_device *ndev,
 {
 	struct axienet_local *lp = netdev_priv(ndev);
 
+	ecoalesce->use_adaptive_rx_coalesce = lp->rx_dim_enabled;
 	ecoalesce->rx_max_coalesced_frames = lp->coalesce_count_rx;
 	ecoalesce->rx_coalesce_usecs = lp->coalesce_usec_rx;
 	ecoalesce->tx_max_coalesced_frames = lp->coalesce_count_tx;
@@ -2127,9 +2169,21 @@ axienet_ethtools_set_coalesce(struct net_device *ndev,
 			      struct netlink_ext_ack *extack)
 {
 	struct axienet_local *lp = netdev_priv(ndev);
+	bool new_dim = ecoalesce->use_adaptive_rx_coalesce;
+	bool old_dim = lp->rx_dim_enabled;
+
+	if (!new_dim) {
+		if (old_dim) {
+			WRITE_ONCE(lp->rx_dim_enabled, false);
+			napi_synchronize(&lp->napi_rx);
+			flush_work(&lp->rx_dim.work);
+		}
+
+		if (ecoalesce->rx_max_coalesced_frames)
+			lp->coalesce_count_rx =
+				ecoalesce->rx_max_coalesced_frames;
+	}
 
-	if (ecoalesce->rx_max_coalesced_frames)
-		lp->coalesce_count_rx = ecoalesce->rx_max_coalesced_frames;
 	if (ecoalesce->rx_coalesce_usecs)
 		lp->coalesce_usec_rx = ecoalesce->rx_coalesce_usecs;
 	if (ecoalesce->tx_max_coalesced_frames)
@@ -2137,6 +2191,11 @@ axienet_ethtools_set_coalesce(struct net_device *ndev,
 	if (ecoalesce->tx_coalesce_usecs)
 		lp->coalesce_usec_tx = ecoalesce->tx_coalesce_usecs;
 
+	if (new_dim && !old_dim) {
+		axienet_dim_coalesce_rx(lp);
+		WRITE_ONCE(lp->rx_dim_enabled, true);
+	}
+
 	axienet_update_coalesce_rx(lp);
 	axienet_update_coalesce_tx(lp);
 	return 0;
@@ -2376,7 +2435,8 @@ axienet_ethtool_get_rmon_stats(struct net_device *dev,
 
 static const struct ethtool_ops axienet_ethtool_ops = {
 	.supported_coalesce_params = ETHTOOL_COALESCE_MAX_FRAMES |
-				     ETHTOOL_COALESCE_USECS,
+				     ETHTOOL_COALESCE_USECS |
+				     ETHTOOL_COALESCE_USE_ADAPTIVE_RX,
 	.get_drvinfo    = axienet_ethtools_get_drvinfo,
 	.get_regs_len   = axienet_ethtools_get_regs_len,
 	.get_regs       = axienet_ethtools_get_regs,
@@ -2925,7 +2985,10 @@ static int axienet_probe(struct platform_device *pdev)
 
 	spin_lock_init(&lp->rx_cr_lock);
 	spin_lock_init(&lp->tx_cr_lock);
-	lp->coalesce_count_rx = XAXIDMA_DFT_RX_THRESHOLD;
+	INIT_WORK(&lp->rx_dim.work, axienet_rx_dim_work);
+	lp->rx_dim_enabled = true;
+	lp->rx_dim.profile_ix = 1;
+	axienet_dim_coalesce_rx(lp);
 	lp->coalesce_count_tx = XAXIDMA_DFT_TX_THRESHOLD;
 	lp->coalesce_usec_rx = XAXIDMA_DFT_RX_USEC;
 	lp->coalesce_usec_tx = XAXIDMA_DFT_TX_USEC;
-- 
2.35.1.1320.gc452695387.dirty

Re: [PATCH net-next 2/2] net: xilinx: axienet: Enable adaptive IRQ coalescing with DIM

[Eric Dumazet]

On Tue, Sep 3, 2024 at 9:25 PM Sean Anderson <sean.anderson@linux.dev> wrote:
>

> +
> +/**
> + * axienet_rx_dim_work() - Adjust RX DIM settings
> + * @work: The work struct
> + */
> +static void axienet_rx_dim_work(struct work_struct *work)
> +{
> +       struct axienet_local *lp =
> +               container_of(work, struct axienet_local, rx_dim.work);
> +
> +       rtnl_lock();

Why do you need rtnl ?

This is very dangerous, because cancel_work_sync(&lp->rx_dim.work)
might deadlock.


> +       axienet_dim_coalesce_rx(lp);
> +       axienet_update_coalesce_rx(lp);
> +       rtnl_unlock();
> +
> +       lp->rx_dim.state = DIM_START_MEASURE;
> +}
>

Re: [PATCH net-next 2/2] net: xilinx: axienet: Enable adaptive IRQ coalescing with DIM

[Sean Anderson]

On 9/4/24 13:04, Eric Dumazet wrote:
> On Tue, Sep 3, 2024 at 9:25 PM Sean Anderson <sean.anderson@linux.dev> wrote:
>>
> 
>> +
>> +/**
>> + * axienet_rx_dim_work() - Adjust RX DIM settings
>> + * @work: The work struct
>> + */
>> +static void axienet_rx_dim_work(struct work_struct *work)
>> +{
>> +       struct axienet_local *lp =
>> +               container_of(work, struct axienet_local, rx_dim.work);
>> +
>> +       rtnl_lock();
> 
> Why do you need rtnl ?

To protect against concurrent modification in axienet_ethtools_set_coalesce.

> This is very dangerous, because cancel_work_sync(&lp->rx_dim.work)
> might deadlock.

Ah, you're right. So maybe I should add a separate mutex for this.

--Sean

>> +       axienet_dim_coalesce_rx(lp);
>> +       axienet_update_coalesce_rx(lp);
>> +       rtnl_unlock();
>> +
>> +       lp->rx_dim.state = DIM_START_MEASURE;
>> +}
>>

Re: [PATCH net-next 0/2] net: xilinx: axienet: Enable adaptive IRQ coalescing with DIM

[Simon Horman]

On Tue, Sep 03, 2024 at 03:25:22PM -0400, Sean Anderson wrote:
> To improve performance without sacrificing latency under low load,
> enable DIM. While I appreciate not having to write the library myself, I
> do think there are many unusual aspects to DIM, as detailed in the last
> patch.
> 
> This series depends on [1].
> 
> [1] https://lore.kernel.org/netdev/20240903180059.4134461-1-sean.anderson@linux.dev/

Hi Sean,

Unfortunately the CI doesn't understand dependencies,
and so it is unable to apply this patchset :(

I would suggest bundling patches for the same driver for net-next
in a single patchset. And in any case, only having one active
at any given time.

-- 
pw-bot: cr

Re: [PATCH net-next 0/2] net: xilinx: axienet: Enable adaptive IRQ coalescing with DIM

[Sean Anderson]

On 9/4/24 12:35, Simon Horman wrote:
> On Tue, Sep 03, 2024 at 03:25:22PM -0400, Sean Anderson wrote:
>> To improve performance without sacrificing latency under low load,
>> enable DIM. While I appreciate not having to write the library myself, I
>> do think there are many unusual aspects to DIM, as detailed in the last
>> patch.
>> 
>> This series depends on [1].
>> 
>> [1] https://lore.kernel.org/netdev/20240903180059.4134461-1-sean.anderson@linux.dev/
> 
> Hi Sean,
> 
> Unfortunately the CI doesn't understand dependencies,
> and so it is unable to apply this patchset :(
> 
> I would suggest bundling patches for the same driver for net-next
> in a single patchset. And in any case, only having one active
> at any given time.
> 

Well, I would normally do so, but that patch is a fix and this series is
an improvement. So that one goes into net and this one goes into net-next.

I've been advised in the past to split up independent patches so they can be
reviewed/applied individually.

--Sean

Re: [PATCH net-next 0/2] net: xilinx: axienet: Enable adaptive IRQ coalescing with DIM

[Simon Horman]

On Thu, Sep 05, 2024 at 10:27:00AM -0400, Sean Anderson wrote:
> On 9/4/24 12:35, Simon Horman wrote:
> > On Tue, Sep 03, 2024 at 03:25:22PM -0400, Sean Anderson wrote:
> >> To improve performance without sacrificing latency under low load,
> >> enable DIM. While I appreciate not having to write the library myself, I
> >> do think there are many unusual aspects to DIM, as detailed in the last
> >> patch.
> >> 
> >> This series depends on [1].
> >> 
> >> [1] https://lore.kernel.org/netdev/20240903180059.4134461-1-sean.anderson@linux.dev/
> > 
> > Hi Sean,
> > 
> > Unfortunately the CI doesn't understand dependencies,
> > and so it is unable to apply this patchset :(
> > 
> > I would suggest bundling patches for the same driver for net-next
> > in a single patchset. And in any case, only having one active
> > at any given time.
> > 
> 
> Well, I would normally do so, but that patch is a fix and this series is
> an improvement. So that one goes into net and this one goes into net-next.
> 
> I've been advised in the past to split up independent patches so they can be
> reviewed/applied individually.

Thanks Sean,

Understood. Given the first point, which I had missed earlier,
I'd would have suggested marking this patch-set as an RFC,
then reposting it once the dependency hits net-next (via net).