Skip to contentSkip to conversation

arm64: dts: imx8mm: Fix the ENET interrupts order

RFC PATCH In review Fabio Estevam 2 participants 5 messages Aug 18, 2020 – Aug 19, 2020

Subsystems (2)
lore
Share & tools
lore

Fabio Estevam

author reviewer
#0 lore raw

RFC PATCH 1/2 arm64: dts: imx8mm: Fix the ENET interrupts order

According to Documentation/devicetree/bindings/net/fsl-fec.txt, when interrupt-names is not passed the following interrupt order is assumed:

__Number of interrupts__ __Default__

	1			"int0"
	2			"int0", "pps"
	3			"int0", "int1", "int2"
	4			"int0", "int1", "int2", "pps"

In the current imx8mm.dtsi this translates to:

- int0 ---> IRQ 118
- int1 ---> IRQ 119
- int2 ---> IRQ 120

However, just like i.MX7, i.MX8MM uses the following ENET irq mapping:

- int0 ---> IRQ 120
- int1 ---> IRQ 118
- int2 ---> IRQ 119

Fix it by passing the interrupt-names property with the correct mapping.

Tested networking on a imx8mm-evk board successfully.

---
Hi Fugang,

Could you please help review this RFC series?

My understanding is that the i.MX8M class of products are derived from i.MX7 from an ENET IRQ mapping perspective. (i.MX8QXP also uses the same i.MX7 mapping by the way). The Reference Manual also seems to indicate the same, but the ENET IRQ naming differs a bit between the i.MX7 and i.MX8MM RM's.

If this is correct, then I plan to also fix i.MX8MQ, i.MX8MN and i.MX8MP dtsi files.

My initial goal was to add the pps irq (patch 2/2), but then I noticed the potential irq mismatch and now it is a two patch series.

Thanks

 arch/arm64/boot/dts/freescale/imx8mm.dtsi | 7 ++++---
 1 file changed, 4 insertions(+), 3 deletions(-)
arch/arm64/boot/dts/freescale/imx8mm.dtsi +4 -3
index aaf6e71101a1..551afc270fb7 100644
@@ -844,9 +844,10 @@
 			fec1: ethernet@30be0000 {
 				compatible = "fsl,imx8mm-fec", "fsl,imx6sx-fec";
 				reg = <0x30be0000 0x10000>;
-				interrupts = <GIC_SPI 118 IRQ_TYPE_LEVEL_HIGH>,
-					     <GIC_SPI 119 IRQ_TYPE_LEVEL_HIGH>,
-					     <GIC_SPI 120 IRQ_TYPE_LEVEL_HIGH>;
+				interrupt-names = "int0", "int1", "int2";
+				interrupts = <GIC_SPI 120 IRQ_TYPE_LEVEL_HIGH>,
+					<GIC_SPI 118 IRQ_TYPE_LEVEL_HIGH>,
+					<GIC_SPI 119 IRQ_TYPE_LEVEL_HIGH>;
 				clocks = <&clk IMX8MM_CLK_ENET1_ROOT>,
 					 <&clk IMX8MM_CLK_ENET1_ROOT>,
 					 <&clk IMX8MM_CLK_ENET_TIMER>,
Signature · 7 lines
2.17.1


_______________________________________________
linux-arm-kernel mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel
author reviewer
#1 lore raw

RFC PATCH 2/2 arm64: dts: imx8mm: Describe the ENET pps interrupt

imx8mm has IRQ 121 associated with the '1588 Timer Interrupt', so add an entry for it.

With this change IRQ 121 can properly increment when PTP applications like ptp4l is used.

---
 arch/arm64/boot/dts/freescale/imx8mm.dtsi | 5 +++--
 1 file changed, 3 insertions(+), 2 deletions(-)
arch/arm64/boot/dts/freescale/imx8mm.dtsi +3 -2
index 551afc270fb7..6f435628824c 100644
@@ -844,10 +844,11 @@
 			fec1: ethernet@30be0000 {
 				compatible = "fsl,imx8mm-fec", "fsl,imx6sx-fec";
 				reg = <0x30be0000 0x10000>;
-				interrupt-names = "int0", "int1", "int2";
+				interrupt-names = "int0", "int1", "int2", "pps";
 				interrupts = <GIC_SPI 120 IRQ_TYPE_LEVEL_HIGH>,
 					<GIC_SPI 118 IRQ_TYPE_LEVEL_HIGH>,
-					<GIC_SPI 119 IRQ_TYPE_LEVEL_HIGH>;
+					<GIC_SPI 119 IRQ_TYPE_LEVEL_HIGH>,
+					<GIC_SPI 121 IRQ_TYPE_LEVEL_HIGH>;
 				clocks = <&clk IMX8MM_CLK_ENET1_ROOT>,
 					 <&clk IMX8MM_CLK_ENET1_ROOT>,
 					 <&clk IMX8MM_CLK_ENET_TIMER>,
Signature · 7 lines
2.17.1


_______________________________________________
linux-arm-kernel mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel
#3 lore raw
In reply to Fabio Estevam

From: Fabio Estevam <[email protected]> Sent: Wednesday, August 19, 2020 5:05 AM

Quoted 10 lines

Hi Fugang,

Could you please help review this RFC series?

My understanding is that the i.MX8M class of products are derived from i.MX7 from an ENET IRQ mapping perspective. (i.MX8QXP also uses the same i.MX7 mapping by the way). The Reference Manual also seems to indicate the same, but the ENET IRQ naming differs a bit between the
i.MX7 and i.MX8MM RM's.

Quoted 7 lines

If this is correct, then I plan to also fix i.MX8MQ, i.MX8MN and i.MX8MP dtsi files.

My initial goal was to add the pps irq (patch 2/2), but then I noticed the potential irq mismatch and now it is a two patch series.

Thanks

It doesn't matter, since there three irq share the same irq handler, and irq handler distinguish Irq by checking register event, so there have no explicit mapping for the three irqs, so we never see problem. But for the fourth irq that is required for the last one, which is for pps, not for ptp4l.

Regards,
Fugang

_______________________________________________
linux-arm-kernel mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel

#2 lore raw

RE: [EXT] [RFC PATCH 2/2] arm64: dts: imx8mm: Describe the ENET pps interrupt

In reply to Fabio Estevam

From: Fabio Estevam <[email protected]> Sent: Wednesday, August 19, 2020 5:05 AM

Quoted 5 lines

imx8mm has IRQ 121 associated with the '1588 Timer Interrupt', so add an entry for it.

With this change IRQ 121 can properly increment when PTP applications like ptp4l is used.

No, IRQ 121 is not for PTP (ptp4l), upstream kernel already support PTP, the irq 121 is only for PPS, not relates to PTP. And PPS is optional that depends on board design and real use case.

Quoted 32 lines
---
 arch/arm64/boot/dts/freescale/imx8mm.dtsi | 5 +++--
 1 file changed, 3 insertions(+), 2 deletions(-)
arch/arm64/boot/dts/freescale/imx8mm.dtsi +2 -2
index 551afc270fb7..6f435628824c 100644
@@ -844,10 +844,11 @@
                        fec1: ethernet@30be0000 {
                                compatible = "fsl,imx8mm-fec",
 "fsl,imx6sx-fec";
                                reg = <0x30be0000 0x10000>;
-                               interrupt-names = "int0", "int1", "int2";
+                               interrupt-names = "int0", "int1",
+ "int2", "pps";
                                interrupts = <GIC_SPI 120
 IRQ_TYPE_LEVEL_HIGH>,
                                        <GIC_SPI 118
 IRQ_TYPE_LEVEL_HIGH>,
-                                       <GIC_SPI 119

IRQ_TYPE_LEVEL_HIGH>;
+ <GIC_SPI 119
IRQ_TYPE_LEVEL_HIGH>,
+ <GIC_SPI 121
+ IRQ_TYPE_LEVEL_HIGH>;
clocks = <&clk
IMX8MM_CLK_ENET1_ROOT>,

Quoted 6 lines
                                         <&clk
IMX8MM_CLK_ENET1_ROOT>,
                                         <&clk

IMX8MM_CLK_ENET_TIMER>,
--
2.17.1

_______________________________________________
linux-arm-kernel mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel

author reviewer
#4 lore raw
In reply to Andy Duan

Hi Andy,

Quoted 3 lines · Andy Duan

It doesn't matter, since there three irq share the same irq handler, and irq handler distinguish Irq by checking register event, so there have no explicit mapping for the three irqs, so we never see problem. But for the fourth irq that is required for the last one, which is for pps, not for ptp4l.

Thanks. I will submit patches adding the fourth interrupt for the i.MX8M SoCs.

_______________________________________________
linux-arm-kernel mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel