[net-next v5 0/3] rework IRQ handling in mtk_eth_soc

[net-next v5 0/3] rework IRQ handling in mtk_eth_soc

[Frank Wunderlich]

From: Frank Wunderlich <frank-w@public-files.de>

This series introduces named IRQs while keeping the index based way
for older dts.
Further it makes some cleanup like adding consts for index access and
avoids loading first IRQ which was not used on non SHARED_INT SoCs.
    
changes:
    
    v5:
    - fixed typo and add linebreak in description of patch 1
    - moved comments from previous patch #3 to patch #1 with changes suggested by simon
    - rename consts to be compatible with upcoming RSS/LRO changes
      MTK_ETH_IRQ_SHARED => MTK_FE_IRQ_SHARED
      MTK_ETH_IRQ_TX => MTK_FE_IRQ_TX
      MTK_ETH_IRQ_RX => MTK_FE_IRQ_RX
      MTK_ETH_IRQ_MAX => MTK_FE_IRQ_NUM
    - change commit title and description in patch 3
    
    v4:
    - calculate max from last (rx) irq index and use it for array size too
    - drop >2 condition as max is already 2 and drop the else continue
    - update comment to explain which IRQs are taken in legacy way
    
    v3:
    added patches
    - #2 (add constants for irq index)
    - #3 (skip first IRQ on ! MTK_SHARED_INT)
    to the v2 non-series patch
    
    https://patchwork.kernel.org/project/netdevbpf/patch/20250615084521.32329-1-linux@fw-web.de/
    
Tested on BPI-R4/mt7988 with IRQ names and BPI-R2/mt7623 and BPI-R3/mt7986 with upstreamed
dts via index-mode.
I do not have any MTK_SHARED_INT (mt7621/mt7628) boards to testing.

Frank Wunderlich (3):
  net: ethernet: mtk_eth_soc: support named IRQs
  net: ethernet: mtk_eth_soc: add consts for irq index
  net: ethernet: mtk_eth_soc: skip first IRQ if not used

 drivers/net/ethernet/mediatek/mtk_eth_soc.c | 62 +++++++++++++++------
 drivers/net/ethernet/mediatek/mtk_eth_soc.h |  7 ++-
 2 files changed, 51 insertions(+), 18 deletions(-)

-- 
2.43.0

[net-next v5 1/3] net: ethernet: mtk_eth_soc: support named IRQs

[Frank Wunderlich]

From: Frank Wunderlich <frank-w@public-files.de>

Add named interrupts and keep index based fallback for existing
devicetrees.

Currently only rx and tx IRQs are defined to be used with mt7988, but
later extended with RSS/LRO support.

Signed-off-by: Frank Wunderlich <frank-w@public-files.de>
Reviewed-by: Simon Horman <horms@kernel.org>
---
v5:
- fix typo in description
- add comments from previous patch #3 with changes suggested by simon

v2:
- move irqs loading part into own helper function
- reduce indentation
- place mtk_get_irqs helper before the irq_handler (note for simon)
---
 drivers/net/ethernet/mediatek/mtk_eth_soc.c | 46 ++++++++++++++++-----
 1 file changed, 35 insertions(+), 11 deletions(-)

diff --git a/drivers/net/ethernet/mediatek/mtk_eth_soc.c b/drivers/net/ethernet/mediatek/mtk_eth_soc.c
index b76d35069887..39b673ed7495 100644
--- a/drivers/net/ethernet/mediatek/mtk_eth_soc.c
+++ b/drivers/net/ethernet/mediatek/mtk_eth_soc.c
@@ -3337,6 +3337,37 @@ static void mtk_tx_timeout(struct net_device *dev, unsigned int txqueue)
 	schedule_work(&eth->pending_work);
 }
 
+static int mtk_get_irqs(struct platform_device *pdev, struct mtk_eth *eth)
+{
+	int i;
+
+	/* future SoCs beginning with MT7988 should use named IRQs in dts */
+	eth->irq[1] = platform_get_irq_byname(pdev, "tx");
+	eth->irq[2] = platform_get_irq_byname(pdev, "rx");
+	if (eth->irq[1] >= 0 && eth->irq[2] >= 0)
+		return 0;
+
+	/* legacy way:
+	 * On MTK_SHARED_INT SoCs (MT7621 + MT7628) the first IRQ is taken
+	 * from devicetree and used for both RX and TX - it is shared.
+	 * On SoCs with non-shared IRQs the first entry is not used,
+	 * the second is for TX, and the third is for RX.
+	 */
+	for (i = 0; i < 3; i++) {
+		if (MTK_HAS_CAPS(eth->soc->caps, MTK_SHARED_INT) && i > 0)
+			eth->irq[i] = eth->irq[0];
+		else
+			eth->irq[i] = platform_get_irq(pdev, i);
+
+		if (eth->irq[i] < 0) {
+			dev_err(&pdev->dev, "no IRQ%d resource found\n", i);
+			return -ENXIO;
+		}
+	}
+
+	return 0;
+}
+
 static irqreturn_t mtk_handle_irq_rx(int irq, void *_eth)
 {
 	struct mtk_eth *eth = _eth;
@@ -5106,17 +5137,10 @@ static int mtk_probe(struct platform_device *pdev)
 		}
 	}
 
-	for (i = 0; i < 3; i++) {
-		if (MTK_HAS_CAPS(eth->soc->caps, MTK_SHARED_INT) && i > 0)
-			eth->irq[i] = eth->irq[0];
-		else
-			eth->irq[i] = platform_get_irq(pdev, i);
-		if (eth->irq[i] < 0) {
-			dev_err(&pdev->dev, "no IRQ%d resource found\n", i);
-			err = -ENXIO;
-			goto err_wed_exit;
-		}
-	}
+	err = mtk_get_irqs(pdev, eth);
+	if (err)
+		goto err_wed_exit;
+
 	for (i = 0; i < ARRAY_SIZE(eth->clks); i++) {
 		eth->clks[i] = devm_clk_get(eth->dev,
 					    mtk_clks_source_name[i]);
-- 
2.43.0

[net-next v5 2/3] net: ethernet: mtk_eth_soc: add consts for irq index

[Frank Wunderlich]

From: Frank Wunderlich <frank-w@public-files.de>

Use consts instead of fixed integers for accessing IRQ array.

Signed-off-by: Frank Wunderlich <frank-w@public-files.de>
Reviewed-by: Simon Horman <horms@kernel.org>
---
v5:
- rename consts to be compatible with upcoming RSS/LRO changes
  MTK_ETH_IRQ_SHARED => MTK_FE_IRQ_SHARED
  MTK_ETH_IRQ_TX => MTK_FE_IRQ_TX
  MTK_ETH_IRQ_RX => MTK_FE_IRQ_RX
  MTK_ETH_IRQ_MAX => MTK_FE_IRQ_NUM
v4:
- calculate max from last (rx) irq index and use it for array size too
---
 drivers/net/ethernet/mediatek/mtk_eth_soc.c | 22 ++++++++++-----------
 drivers/net/ethernet/mediatek/mtk_eth_soc.h |  7 ++++++-
 2 files changed, 17 insertions(+), 12 deletions(-)

diff --git a/drivers/net/ethernet/mediatek/mtk_eth_soc.c b/drivers/net/ethernet/mediatek/mtk_eth_soc.c
index 39b673ed7495..875e477a987b 100644
--- a/drivers/net/ethernet/mediatek/mtk_eth_soc.c
+++ b/drivers/net/ethernet/mediatek/mtk_eth_soc.c
@@ -3342,9 +3342,9 @@ static int mtk_get_irqs(struct platform_device *pdev, struct mtk_eth *eth)
 	int i;
 
 	/* future SoCs beginning with MT7988 should use named IRQs in dts */
-	eth->irq[1] = platform_get_irq_byname(pdev, "tx");
-	eth->irq[2] = platform_get_irq_byname(pdev, "rx");
-	if (eth->irq[1] >= 0 && eth->irq[2] >= 0)
+	eth->irq[MTK_FE_IRQ_TX] = platform_get_irq_byname(pdev, "tx");
+	eth->irq[MTK_FE_IRQ_RX] = platform_get_irq_byname(pdev, "rx");
+	if (eth->irq[MTK_FE_IRQ_TX] >= 0 && eth->irq[MTK_FE_IRQ_RX] >= 0)
 		return 0;
 
 	/* legacy way:
@@ -3353,9 +3353,9 @@ static int mtk_get_irqs(struct platform_device *pdev, struct mtk_eth *eth)
 	 * On SoCs with non-shared IRQs the first entry is not used,
 	 * the second is for TX, and the third is for RX.
 	 */
-	for (i = 0; i < 3; i++) {
+	for (i = 0; i < MTK_FE_IRQ_NUM; i++) {
 		if (MTK_HAS_CAPS(eth->soc->caps, MTK_SHARED_INT) && i > 0)
-			eth->irq[i] = eth->irq[0];
+			eth->irq[i] = eth->irq[MTK_FE_IRQ_SHARED];
 		else
 			eth->irq[i] = platform_get_irq(pdev, i);
 
@@ -3421,7 +3421,7 @@ static void mtk_poll_controller(struct net_device *dev)
 
 	mtk_tx_irq_disable(eth, MTK_TX_DONE_INT);
 	mtk_rx_irq_disable(eth, eth->soc->rx.irq_done_mask);
-	mtk_handle_irq_rx(eth->irq[2], dev);
+	mtk_handle_irq_rx(eth->irq[MTK_FE_IRQ_RX], dev);
 	mtk_tx_irq_enable(eth, MTK_TX_DONE_INT);
 	mtk_rx_irq_enable(eth, eth->soc->rx.irq_done_mask);
 }
@@ -4907,7 +4907,7 @@ static int mtk_add_mac(struct mtk_eth *eth, struct device_node *np)
 	eth->netdev[id]->features |= eth->soc->hw_features;
 	eth->netdev[id]->ethtool_ops = &mtk_ethtool_ops;
 
-	eth->netdev[id]->irq = eth->irq[0];
+	eth->netdev[id]->irq = eth->irq[MTK_FE_IRQ_SHARED];
 	eth->netdev[id]->dev.of_node = np;
 
 	if (MTK_HAS_CAPS(eth->soc->caps, MTK_SOC_MT7628))
@@ -5184,17 +5184,17 @@ static int mtk_probe(struct platform_device *pdev)
 	}
 
 	if (MTK_HAS_CAPS(eth->soc->caps, MTK_SHARED_INT)) {
-		err = devm_request_irq(eth->dev, eth->irq[0],
+		err = devm_request_irq(eth->dev, eth->irq[MTK_FE_IRQ_SHARED],
 				       mtk_handle_irq, 0,
 				       dev_name(eth->dev), eth);
 	} else {
-		err = devm_request_irq(eth->dev, eth->irq[1],
+		err = devm_request_irq(eth->dev, eth->irq[MTK_FE_IRQ_TX],
 				       mtk_handle_irq_tx, 0,
 				       dev_name(eth->dev), eth);
 		if (err)
 			goto err_free_dev;
 
-		err = devm_request_irq(eth->dev, eth->irq[2],
+		err = devm_request_irq(eth->dev, eth->irq[MTK_FE_IRQ_RX],
 				       mtk_handle_irq_rx, 0,
 				       dev_name(eth->dev), eth);
 	}
@@ -5240,7 +5240,7 @@ static int mtk_probe(struct platform_device *pdev)
 		} else
 			netif_info(eth, probe, eth->netdev[i],
 				   "mediatek frame engine at 0x%08lx, irq %d\n",
-				   eth->netdev[i]->base_addr, eth->irq[0]);
+				   eth->netdev[i]->base_addr, eth->irq[MTK_FE_IRQ_SHARED]);
 	}
 
 	/* we run 2 devices on the same DMA ring so we need a dummy device
diff --git a/drivers/net/ethernet/mediatek/mtk_eth_soc.h b/drivers/net/ethernet/mediatek/mtk_eth_soc.h
index 6f72a8c8ae1e..8cdf1317dff5 100644
--- a/drivers/net/ethernet/mediatek/mtk_eth_soc.h
+++ b/drivers/net/ethernet/mediatek/mtk_eth_soc.h
@@ -642,6 +642,11 @@
 
 #define MTK_MAC_FSM(x)		(0x1010C + ((x) * 0x100))
 
+#define MTK_FE_IRQ_SHARED	0
+#define MTK_FE_IRQ_TX		1
+#define MTK_FE_IRQ_RX		2
+#define MTK_FE_IRQ_NUM		(MTK_FE_IRQ_RX + 1)
+
 struct mtk_rx_dma {
 	unsigned int rxd1;
 	unsigned int rxd2;
@@ -1292,7 +1297,7 @@ struct mtk_eth {
 	struct net_device		*dummy_dev;
 	struct net_device		*netdev[MTK_MAX_DEVS];
 	struct mtk_mac			*mac[MTK_MAX_DEVS];
-	int				irq[3];
+	int				irq[MTK_FE_IRQ_NUM];
 	u32				msg_enable;
 	unsigned long			sysclk;
 	struct regmap			*ethsys;
-- 
2.43.0

[net-next v5 3/3] net: ethernet: mtk_eth_soc: skip first IRQ if not used

[Frank Wunderlich]

From: Frank Wunderlich <frank-w@public-files.de>

On SoCs without MTK_SHARED_INT capability (all except mt7621 and
mt7628) platform_get_irq() is called for the first IRQ (eth->irq[0])
but it is never used.
Skip the first IRQ and reduce the IRQ-count to 2.

Signed-off-by: Frank Wunderlich <frank-w@public-files.de>
---
v5:
- change commit title and description
v4:
- drop >2 condition as max is already 2 and drop the else continue
- update comment to explain which IRQs are taken in legacy way
---
 drivers/net/ethernet/mediatek/mtk_eth_soc.c | 12 ++++++++----
 drivers/net/ethernet/mediatek/mtk_eth_soc.h |  4 ++--
 2 files changed, 10 insertions(+), 6 deletions(-)

diff --git a/drivers/net/ethernet/mediatek/mtk_eth_soc.c b/drivers/net/ethernet/mediatek/mtk_eth_soc.c
index 875e477a987b..7990c84b2b56 100644
--- a/drivers/net/ethernet/mediatek/mtk_eth_soc.c
+++ b/drivers/net/ethernet/mediatek/mtk_eth_soc.c
@@ -3354,10 +3354,14 @@ static int mtk_get_irqs(struct platform_device *pdev, struct mtk_eth *eth)
 	 * the second is for TX, and the third is for RX.
 	 */
 	for (i = 0; i < MTK_FE_IRQ_NUM; i++) {
-		if (MTK_HAS_CAPS(eth->soc->caps, MTK_SHARED_INT) && i > 0)
-			eth->irq[i] = eth->irq[MTK_FE_IRQ_SHARED];
-		else
-			eth->irq[i] = platform_get_irq(pdev, i);
+		if (MTK_HAS_CAPS(eth->soc->caps, MTK_SHARED_INT)) {
+			if (i == 0)
+				eth->irq[MTK_FE_IRQ_SHARED] = platform_get_irq(pdev, i);
+			else
+				eth->irq[i] = eth->irq[MTK_FE_IRQ_SHARED];
+		} else {
+			eth->irq[i] = platform_get_irq(pdev, i + 1);
+		}
 
 		if (eth->irq[i] < 0) {
 			dev_err(&pdev->dev, "no IRQ%d resource found\n", i);
diff --git a/drivers/net/ethernet/mediatek/mtk_eth_soc.h b/drivers/net/ethernet/mediatek/mtk_eth_soc.h
index 8cdf1317dff5..9261c0e13b59 100644
--- a/drivers/net/ethernet/mediatek/mtk_eth_soc.h
+++ b/drivers/net/ethernet/mediatek/mtk_eth_soc.h
@@ -643,8 +643,8 @@
 #define MTK_MAC_FSM(x)		(0x1010C + ((x) * 0x100))
 
 #define MTK_FE_IRQ_SHARED	0
-#define MTK_FE_IRQ_TX		1
-#define MTK_FE_IRQ_RX		2
+#define MTK_FE_IRQ_TX		0
+#define MTK_FE_IRQ_RX		1
 #define MTK_FE_IRQ_NUM		(MTK_FE_IRQ_RX + 1)
 
 struct mtk_rx_dma {
-- 
2.43.0

Re: [net-next v5 1/3] net: ethernet: mtk_eth_soc: support named IRQs

[Daniel Golle]

On Wed, Jun 18, 2025 at 03:07:12PM +0200, Frank Wunderlich wrote:
> From: Frank Wunderlich <frank-w@public-files.de>
> 
> Add named interrupts and keep index based fallback for existing
> devicetrees.
> 
> Currently only rx and tx IRQs are defined to be used with mt7988, but
> later extended with RSS/LRO support.
> 
> Signed-off-by: Frank Wunderlich <frank-w@public-files.de>
> Reviewed-by: Simon Horman <horms@kernel.org>
>  
> +static int mtk_get_irqs(struct platform_device *pdev, struct mtk_eth *eth)
> +{
> +	int i;
> +
> +	/* future SoCs beginning with MT7988 should use named IRQs in dts */
> +	eth->irq[1] = platform_get_irq_byname(pdev, "tx");
> +	eth->irq[2] = platform_get_irq_byname(pdev, "rx");
> +	if (eth->irq[1] >= 0 && eth->irq[2] >= 0)
> +		return 0;

I'd rather extend that logic and fall back to the legacy way only in case
of -ENXIO. Ie. add here:

if (eth->irq[1] != -ENXIO)
	return eth->irq[1];

if (eth->irq[2] != -ENXIO)
	return eth->irq[2];

Maybe also output a warning at this point in case MTK_SHARED_INT is no
set, to recommend users to update their device tree to named interrupts.

> +
> +	/* legacy way:
> +	 * On MTK_SHARED_INT SoCs (MT7621 + MT7628) the first IRQ is taken
> +	 * from devicetree and used for both RX and TX - it is shared.
> +	 * On SoCs with non-shared IRQs the first entry is not used,
> +	 * the second is for TX, and the third is for RX.
> +	 */
> +	for (i = 0; i < 3; i++) {
> +		if (MTK_HAS_CAPS(eth->soc->caps, MTK_SHARED_INT) && i > 0)
> +			eth->irq[i] = eth->irq[0];
> +		else
> +			eth->irq[i] = platform_get_irq(pdev, i);
> +
> +		if (eth->irq[i] < 0) {
> +			dev_err(&pdev->dev, "no IRQ%d resource found\n", i);
> +			return -ENXIO;
> +		}
> +	}
> +
> +	return 0;
> +}
> +
>  static irqreturn_t mtk_handle_irq_rx(int irq, void *_eth)
>  {
>  	struct mtk_eth *eth = _eth;
> @@ -5106,17 +5137,10 @@ static int mtk_probe(struct platform_device *pdev)
>  		}
>  	}
>  
> -	for (i = 0; i < 3; i++) {
> -		if (MTK_HAS_CAPS(eth->soc->caps, MTK_SHARED_INT) && i > 0)
> -			eth->irq[i] = eth->irq[0];
> -		else
> -			eth->irq[i] = platform_get_irq(pdev, i);
> -		if (eth->irq[i] < 0) {
> -			dev_err(&pdev->dev, "no IRQ%d resource found\n", i);
> -			err = -ENXIO;
> -			goto err_wed_exit;
> -		}
> -	}
> +	err = mtk_get_irqs(pdev, eth);
> +	if (err)
> +		goto err_wed_exit;
> +
>  	for (i = 0; i < ARRAY_SIZE(eth->clks); i++) {
>  		eth->clks[i] = devm_clk_get(eth->dev,
>  					    mtk_clks_source_name[i]);
> -- 
> 2.43.0
> 
> 

Re: [net-next v5 2/3] net: ethernet: mtk_eth_soc: add consts for irq index

[Daniel Golle]

On Wed, Jun 18, 2025 at 03:07:13PM +0200, Frank Wunderlich wrote:
> From: Frank Wunderlich <frank-w@public-files.de>
> 
> Use consts instead of fixed integers for accessing IRQ array.
> 
> Signed-off-by: Frank Wunderlich <frank-w@public-files.de>
> Reviewed-by: Simon Horman <horms@kernel.org>

Reviewed-by: Daniel Golle <daniel@makrotopia.org>

> ---
> v5:
> - rename consts to be compatible with upcoming RSS/LRO changes
>   MTK_ETH_IRQ_SHARED => MTK_FE_IRQ_SHARED
>   MTK_ETH_IRQ_TX => MTK_FE_IRQ_TX
>   MTK_ETH_IRQ_RX => MTK_FE_IRQ_RX
>   MTK_ETH_IRQ_MAX => MTK_FE_IRQ_NUM
> v4:
> - calculate max from last (rx) irq index and use it for array size too
> ---
>  drivers/net/ethernet/mediatek/mtk_eth_soc.c | 22 ++++++++++-----------
>  drivers/net/ethernet/mediatek/mtk_eth_soc.h |  7 ++++++-
>  2 files changed, 17 insertions(+), 12 deletions(-)
> 
> diff --git a/drivers/net/ethernet/mediatek/mtk_eth_soc.c b/drivers/net/ethernet/mediatek/mtk_eth_soc.c
> index 39b673ed7495..875e477a987b 100644
> --- a/drivers/net/ethernet/mediatek/mtk_eth_soc.c
> +++ b/drivers/net/ethernet/mediatek/mtk_eth_soc.c
> @@ -3342,9 +3342,9 @@ static int mtk_get_irqs(struct platform_device *pdev, struct mtk_eth *eth)
>  	int i;
>  
>  	/* future SoCs beginning with MT7988 should use named IRQs in dts */
> -	eth->irq[1] = platform_get_irq_byname(pdev, "tx");
> -	eth->irq[2] = platform_get_irq_byname(pdev, "rx");
> -	if (eth->irq[1] >= 0 && eth->irq[2] >= 0)
> +	eth->irq[MTK_FE_IRQ_TX] = platform_get_irq_byname(pdev, "tx");
> +	eth->irq[MTK_FE_IRQ_RX] = platform_get_irq_byname(pdev, "rx");
> +	if (eth->irq[MTK_FE_IRQ_TX] >= 0 && eth->irq[MTK_FE_IRQ_RX] >= 0)
>  		return 0;
>  
>  	/* legacy way:
> @@ -3353,9 +3353,9 @@ static int mtk_get_irqs(struct platform_device *pdev, struct mtk_eth *eth)
>  	 * On SoCs with non-shared IRQs the first entry is not used,
>  	 * the second is for TX, and the third is for RX.
>  	 */
> -	for (i = 0; i < 3; i++) {
> +	for (i = 0; i < MTK_FE_IRQ_NUM; i++) {
>  		if (MTK_HAS_CAPS(eth->soc->caps, MTK_SHARED_INT) && i > 0)
> -			eth->irq[i] = eth->irq[0];
> +			eth->irq[i] = eth->irq[MTK_FE_IRQ_SHARED];
>  		else
>  			eth->irq[i] = platform_get_irq(pdev, i);
>  
> @@ -3421,7 +3421,7 @@ static void mtk_poll_controller(struct net_device *dev)
>  
>  	mtk_tx_irq_disable(eth, MTK_TX_DONE_INT);
>  	mtk_rx_irq_disable(eth, eth->soc->rx.irq_done_mask);
> -	mtk_handle_irq_rx(eth->irq[2], dev);
> +	mtk_handle_irq_rx(eth->irq[MTK_FE_IRQ_RX], dev);
>  	mtk_tx_irq_enable(eth, MTK_TX_DONE_INT);
>  	mtk_rx_irq_enable(eth, eth->soc->rx.irq_done_mask);
>  }
> @@ -4907,7 +4907,7 @@ static int mtk_add_mac(struct mtk_eth *eth, struct device_node *np)
>  	eth->netdev[id]->features |= eth->soc->hw_features;
>  	eth->netdev[id]->ethtool_ops = &mtk_ethtool_ops;
>  
> -	eth->netdev[id]->irq = eth->irq[0];
> +	eth->netdev[id]->irq = eth->irq[MTK_FE_IRQ_SHARED];
>  	eth->netdev[id]->dev.of_node = np;
>  
>  	if (MTK_HAS_CAPS(eth->soc->caps, MTK_SOC_MT7628))
> @@ -5184,17 +5184,17 @@ static int mtk_probe(struct platform_device *pdev)
>  	}
>  
>  	if (MTK_HAS_CAPS(eth->soc->caps, MTK_SHARED_INT)) {
> -		err = devm_request_irq(eth->dev, eth->irq[0],
> +		err = devm_request_irq(eth->dev, eth->irq[MTK_FE_IRQ_SHARED],
>  				       mtk_handle_irq, 0,
>  				       dev_name(eth->dev), eth);
>  	} else {
> -		err = devm_request_irq(eth->dev, eth->irq[1],
> +		err = devm_request_irq(eth->dev, eth->irq[MTK_FE_IRQ_TX],
>  				       mtk_handle_irq_tx, 0,
>  				       dev_name(eth->dev), eth);
>  		if (err)
>  			goto err_free_dev;
>  
> -		err = devm_request_irq(eth->dev, eth->irq[2],
> +		err = devm_request_irq(eth->dev, eth->irq[MTK_FE_IRQ_RX],
>  				       mtk_handle_irq_rx, 0,
>  				       dev_name(eth->dev), eth);
>  	}
> @@ -5240,7 +5240,7 @@ static int mtk_probe(struct platform_device *pdev)
>  		} else
>  			netif_info(eth, probe, eth->netdev[i],
>  				   "mediatek frame engine at 0x%08lx, irq %d\n",
> -				   eth->netdev[i]->base_addr, eth->irq[0]);
> +				   eth->netdev[i]->base_addr, eth->irq[MTK_FE_IRQ_SHARED]);
>  	}
>  
>  	/* we run 2 devices on the same DMA ring so we need a dummy device
> diff --git a/drivers/net/ethernet/mediatek/mtk_eth_soc.h b/drivers/net/ethernet/mediatek/mtk_eth_soc.h
> index 6f72a8c8ae1e..8cdf1317dff5 100644
> --- a/drivers/net/ethernet/mediatek/mtk_eth_soc.h
> +++ b/drivers/net/ethernet/mediatek/mtk_eth_soc.h
> @@ -642,6 +642,11 @@
>  
>  #define MTK_MAC_FSM(x)		(0x1010C + ((x) * 0x100))
>  
> +#define MTK_FE_IRQ_SHARED	0
> +#define MTK_FE_IRQ_TX		1
> +#define MTK_FE_IRQ_RX		2
> +#define MTK_FE_IRQ_NUM		(MTK_FE_IRQ_RX + 1)
> +
>  struct mtk_rx_dma {
>  	unsigned int rxd1;
>  	unsigned int rxd2;
> @@ -1292,7 +1297,7 @@ struct mtk_eth {
>  	struct net_device		*dummy_dev;
>  	struct net_device		*netdev[MTK_MAX_DEVS];
>  	struct mtk_mac			*mac[MTK_MAX_DEVS];
> -	int				irq[3];
> +	int				irq[MTK_FE_IRQ_NUM];
>  	u32				msg_enable;
>  	unsigned long			sysclk;
>  	struct regmap			*ethsys;
> -- 
> 2.43.0
> 

Re: [net-next v5 1/3] net: ethernet: mtk_eth_soc: support named IRQs

[Frank Wunderlich]

Am 18. Juni 2025 15:55:45 MESZ schrieb Daniel Golle <daniel@makrotopia.org>:
>On Wed, Jun 18, 2025 at 03:07:12PM +0200, Frank Wunderlich wrote:
>> From: Frank Wunderlich <frank-w@public-files.de>
>> 
>> Add named interrupts and keep index based fallback for existing
>> devicetrees.
>> 
>> Currently only rx and tx IRQs are defined to be used with mt7988, but
>> later extended with RSS/LRO support.
>> 
>> Signed-off-by: Frank Wunderlich <frank-w@public-files.de>
>> Reviewed-by: Simon Horman <horms@kernel.org>
>>  
>> +static int mtk_get_irqs(struct platform_device *pdev, struct mtk_eth *eth)
>> +{
>> +	int i;
>> +
>> +	/* future SoCs beginning with MT7988 should use named IRQs in dts */
>> +	eth->irq[1] = platform_get_irq_byname(pdev, "tx");
>> +	eth->irq[2] = platform_get_irq_byname(pdev, "rx");
>> +	if (eth->irq[1] >= 0 && eth->irq[2] >= 0)
>> +		return 0;
>
>I'd rather extend that logic and fall back to the legacy way only in case
>of -ENXIO. Ie. add here:
>
>if (eth->irq[1] != -ENXIO)
>	return eth->irq[1];
>
>if (eth->irq[2] != -ENXIO)
>	return eth->irq[2];

I would do this later after the consts are used
 instead of index numbers,just to not add lines
 that are changed later again.Better adding the
 lines already with the consts.

>Maybe also output a warning at this point in case MTK_SHARED_INT is no
>set, to recommend users to update their device tree to named interrupts.

I understand the reason behind (documentation
 which irq is used for which purpose),but
 previous devicetrees of non-shared SoCs using
 at least the reserved irq 0. Mt7986 has 0 and 3
 defined in dts. That could be tricky in binding,
 so my way was starting with irq names now for
 new additions and leaving existing dts as they
 are.

Maybe i should add the mt7988 ethernet binding change from my dts series here...

regards Frank

Re: [net-next v5 3/3] net: ethernet: mtk_eth_soc: skip first IRQ if not used

[Daniel Golle]

On Wed, Jun 18, 2025 at 03:07:14PM +0200, Frank Wunderlich wrote:
> From: Frank Wunderlich <frank-w@public-files.de>
> 
> On SoCs without MTK_SHARED_INT capability (all except mt7621 and
> mt7628) platform_get_irq() is called for the first IRQ (eth->irq[0])
> but it is never used.

I know that technically MTK_SHARED_INT is a capability flag, but it's
rather a non-capability. Hardware having dedicated interrupts for RX and
TX is "more capable" than (older, legacy) hardware with just one shared
interrupt for both...

So maybe better:
"On SoCs with dedicated RX and TX interrupts (all except MT7621 and
MT7628) ..."

Reading the datasheet of some recent MediaTek SoC it is worth
noting that there are 4 interrupts assigned to the frame engine and
the FE_INT_GRP register can be used to assign functions to them.

So technically, calling them RX and TX in DT is wrong, becaues they
are fe_int0, fe_int1, fe_int2 and fe_int3, which are then assigned
one or more functions by the driver using that FE_INT_GRP register.

However, it's the driver then assigns QDMA TX to fe_int1 and RX to
fe_int2 while leaving fe_int0 and fe_int3 unsued.

That's what the magic value 0x21021000 which is written to FE_INT_GRP
register does.

On MT7988 and newer, in addition to those 4 frame engine interrupts
there are **another 4** interrupts for PDMA, typically used to
service 4 RX rings while one of the fe_int* is used to indicate
TX done.

> Skip the first IRQ and reduce the IRQ-count to 2.
> 
> Signed-off-by: Frank Wunderlich <frank-w@public-files.de>
> ---
> v5:
> - change commit title and description
> v4:
> - drop >2 condition as max is already 2 and drop the else continue
> - update comment to explain which IRQs are taken in legacy way
> ---
>  drivers/net/ethernet/mediatek/mtk_eth_soc.c | 12 ++++++++----
>  drivers/net/ethernet/mediatek/mtk_eth_soc.h |  4 ++--
>  2 files changed, 10 insertions(+), 6 deletions(-)
> 
> diff --git a/drivers/net/ethernet/mediatek/mtk_eth_soc.c b/drivers/net/ethernet/mediatek/mtk_eth_soc.c
> index 875e477a987b..7990c84b2b56 100644
> --- a/drivers/net/ethernet/mediatek/mtk_eth_soc.c
> +++ b/drivers/net/ethernet/mediatek/mtk_eth_soc.c
> @@ -3354,10 +3354,14 @@ static int mtk_get_irqs(struct platform_device *pdev, struct mtk_eth *eth)
>  	 * the second is for TX, and the third is for RX.
>  	 */
>  	for (i = 0; i < MTK_FE_IRQ_NUM; i++) {
> -		if (MTK_HAS_CAPS(eth->soc->caps, MTK_SHARED_INT) && i > 0)
> -			eth->irq[i] = eth->irq[MTK_FE_IRQ_SHARED];
> -		else
> -			eth->irq[i] = platform_get_irq(pdev, i);
> +		if (MTK_HAS_CAPS(eth->soc->caps, MTK_SHARED_INT)) {
> +			if (i == 0)

This would make it even more readable:
			if (i == MTK_FE_IRQ_SHARED)


Other than that looks good to me:

Reviewed-by: Daniel Golle <daniel@makrotopia.org>

> +				eth->irq[MTK_FE_IRQ_SHARED] = platform_get_irq(pdev, i);
> +			else
> +				eth->irq[i] = eth->irq[MTK_FE_IRQ_SHARED];
> +		} else {
> +			eth->irq[i] = platform_get_irq(pdev, i + 1);
> +		}
>  
>  		if (eth->irq[i] < 0) {
>  			dev_err(&pdev->dev, "no IRQ%d resource found\n", i);
> diff --git a/drivers/net/ethernet/mediatek/mtk_eth_soc.h b/drivers/net/ethernet/mediatek/mtk_eth_soc.h
> index 8cdf1317dff5..9261c0e13b59 100644
> --- a/drivers/net/ethernet/mediatek/mtk_eth_soc.h
> +++ b/drivers/net/ethernet/mediatek/mtk_eth_soc.h
> @@ -643,8 +643,8 @@
>  #define MTK_MAC_FSM(x)		(0x1010C + ((x) * 0x100))
>  
>  #define MTK_FE_IRQ_SHARED	0
> -#define MTK_FE_IRQ_TX		1
> -#define MTK_FE_IRQ_RX		2
> +#define MTK_FE_IRQ_TX		0
> +#define MTK_FE_IRQ_RX		1
>  #define MTK_FE_IRQ_NUM		(MTK_FE_IRQ_RX + 1)
>  
>  struct mtk_rx_dma {
> -- 
> 2.43.0
> 
> 

Re: [net-next v5 3/3] net: ethernet: mtk_eth_soc: skip first IRQ if not used

[Simon Horman]

On Wed, Jun 18, 2025 at 03:07:14PM +0200, Frank Wunderlich wrote:
> From: Frank Wunderlich <frank-w@public-files.de>
> 
> On SoCs without MTK_SHARED_INT capability (all except mt7621 and
> mt7628) platform_get_irq() is called for the first IRQ (eth->irq[0])
> but it is never used.
> Skip the first IRQ and reduce the IRQ-count to 2.
> 
> Signed-off-by: Frank Wunderlich <frank-w@public-files.de>
> ---
> v5:
> - change commit title and description
> v4:
> - drop >2 condition as max is already 2 and drop the else continue
> - update comment to explain which IRQs are taken in legacy way

Reviewed-by: Simon Horman <horms@kernel.org>

Aw: Re: [net-next v5 3/3] net: ethernet: mtk_eth_soc: skip first IRQ if not used

[Frank Wunderlich]

Hi Simon

> Gesendet: Donnerstag, 19. Juni 2025 um 12:03
> Von: "Simon Horman" <horms@kernel.org>
> Betreff: Re: [net-next v5 3/3] net: ethernet: mtk_eth_soc: skip first IRQ if not used
>
> On Wed, Jun 18, 2025 at 03:07:14PM +0200, Frank Wunderlich wrote:
> > From: Frank Wunderlich <frank-w@public-files.de>
> > 
> > On SoCs without MTK_SHARED_INT capability (all except mt7621 and
> > mt7628) platform_get_irq() is called for the first IRQ (eth->irq[0])
> > but it is never used.
> > Skip the first IRQ and reduce the IRQ-count to 2.
> > 
> > Signed-off-by: Frank Wunderlich <frank-w@public-files.de>
> > ---
> > v5:
> > - change commit title and description
> > v4:
> > - drop >2 condition as max is already 2 and drop the else continue
> > - update comment to explain which IRQs are taken in legacy way
> 
> Reviewed-by: Simon Horman <horms@kernel.org>

thank you very much

i guess RB is still valid with changes requested by Daniel?

    net: ethernet: mtk_eth_soc: skip first IRQ if not used
    
    On SoCs with dedicated RX and TX interrupts (all except MT7621 and
    MT7628) platform_get_irq() is called for the first IRQ (eth->irq[0])
    but it is never used.
    Skip the first IRQ and reduce the IRQ-count to 2.

code (changed condition "if (i == MTK_FE_IRQ_SHARED)")
        for (i = 0; i < MTK_FE_IRQ_NUM; i++) {
-               if (MTK_HAS_CAPS(eth->soc->caps, MTK_SHARED_INT) && i > 0)
-                       eth->irq[i] = eth->irq[MTK_FE_IRQ_SHARED];
-               else
-                       eth->irq[i] = platform_get_irq(pdev, i);
+               if (MTK_HAS_CAPS(eth->soc->caps, MTK_SHARED_INT)) {
+                       if (i == MTK_FE_IRQ_SHARED)
+                               eth->irq[MTK_FE_IRQ_SHARED] = platform_get_irq(pdev, i);
+                       else
+                               eth->irq[i] = eth->irq[MTK_FE_IRQ_SHARED];
+               } else {
+                       eth->irq[i] = platform_get_irq(pdev, i + 1);
+               }

regards Frank

Re: Re: [net-next v5 3/3] net: ethernet: mtk_eth_soc: skip first IRQ if not used

[Simon Horman]

On Thu, Jun 19, 2025 at 10:32:05AM +0000, Frank Wunderlich wrote:
> Hi Simon
> 
> > Gesendet: Donnerstag, 19. Juni 2025 um 12:03
> > Von: "Simon Horman" <horms@kernel.org>
> > Betreff: Re: [net-next v5 3/3] net: ethernet: mtk_eth_soc: skip first IRQ if not used
> >
> > On Wed, Jun 18, 2025 at 03:07:14PM +0200, Frank Wunderlich wrote:
> > > From: Frank Wunderlich <frank-w@public-files.de>
> > > 
> > > On SoCs without MTK_SHARED_INT capability (all except mt7621 and
> > > mt7628) platform_get_irq() is called for the first IRQ (eth->irq[0])
> > > but it is never used.
> > > Skip the first IRQ and reduce the IRQ-count to 2.
> > > 
> > > Signed-off-by: Frank Wunderlich <frank-w@public-files.de>
> > > ---
> > > v5:
> > > - change commit title and description
> > > v4:
> > > - drop >2 condition as max is already 2 and drop the else continue
> > > - update comment to explain which IRQs are taken in legacy way
> > 
> > Reviewed-by: Simon Horman <horms@kernel.org>
> 
> thank you very much
> 
> i guess RB is still valid with changes requested by Daniel?

Yes, thanks for asking.