Skip to contentSkip to conversation

ARM: dts: ls1021a-tsn: Use interrupts for the SGMII PHYs

PATCH In review Vladimir Oltean 3 participants 8 messages Nov 9, 2019 1 message not in the archive

Subsystems (3)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 ARM: dts: ls1021a-tsn: Use interrupts for the SGMII PHYs

On the LS1021A-TSN board, the 2 Atheros AR8031 PHYs for eth0 and eth1 have interrupt lines connected to the shared IRQ2_B LS1021A pin.

The interrupts are active low, but the GICv2 controller does not support active-low and falling-edge interrupts, so the only mode it can be configured in is rising-edge.

The interrupt number was obtained by subtracting 32 from the listed interrupt ID from LS1021ARM.pdf Table 5-1. Interrupt assignments.

Switching to interrupts offloads the PHY library from the task of polling the MDIO status and AN registers (1, 4, 5) every second.

Unfortunately, the BCM5464R quad PHY connected to the switch does not appear to have an interrupt line routed to the SoC.

---
 arch/arm/boot/dts/ls1021a-tsn.dts | 4 ++++
 1 file changed, 4 insertions(+)
arch/arm/boot/dts/ls1021a-tsn.dts +4 -0
index 5b7689094b70..4532b2bd3fd1 100644
@@ -203,11 +203,15 @@
 	/* AR8031 */
 	sgmii_phy1: ethernet-phy@1 {
 		reg = <0x1>;
+		/* SGMII1_PHY_INT_B: connected to IRQ2, active low */
+		interrupts = <GIC_SPI 165 IRQ_TYPE_EDGE_RISING>;
 	};
 
 	/* AR8031 */
 	sgmii_phy2: ethernet-phy@2 {
 		reg = <0x2>;
+		/* SGMII2_PHY_INT_B: connected to IRQ2, active low */
+		interrupts = <GIC_SPI 165 IRQ_TYPE_EDGE_RISING>;
 	};
 
 	/* BCM5464 quad PHY */
Signature · 1 line
2.17.1
#1 lore raw
Quoted 6 lines · Vladimir Oltean

On the LS1021A-TSN board, the 2 Atheros AR8031 PHYs for eth0 and eth1 have interrupt lines connected to the shared IRQ2_B LS1021A pin.

The interrupts are active low, but the GICv2 controller does not support active-low and falling-edge interrupts, so the only mode it can be configured in is rising-edge.

Hi Vladimir

So how does this work? The rising edge would occur after the interrupt handler has completed? What triggers the interrupt handler?

Andrew

#2 lore raw
Quoted 8 lines · Andrew Lunn
Quoted 6 lines · Vladimir Oltean

On the LS1021A-TSN board, the 2 Atheros AR8031 PHYs for eth0 and eth1 have interrupt lines connected to the shared IRQ2_B LS1021A pin.

The interrupts are active low, but the GICv2 controller does not support active-low and falling-edge interrupts, so the only mode it can be configured in is rising-edge.

Quoted 7 lines

Hi Vladimir

So how does this work? The rising edge would occur after the interrupt handler has completed? What triggers the interrupt handler?

Andrew

Hi Andrew,

I hope I am not terribly confused about this. I thought I am telling the interrupt controller to raise an IRQ as a result of the low-to-high transition of the electrical signal. Experimentation sure seems to agree with me. So the IRQ is generated immediately _after_ the PHY has left the line in open drain and it got pulled up to Vdd.

Thanks,
-Vladimir

[Sorry for the repost, for some reason Gmail decided to send this email as html earlier]

#8 lore raw
In reply to a message not in the archive
Quoted 17 lines · Vladimir Oltean
Quoted 8 lines · Andrew Lunn
Quoted 6 lines · Vladimir Oltean

On the LS1021A-TSN board, the 2 Atheros AR8031 PHYs for eth0 and eth1 have interrupt lines connected to the shared IRQ2_B LS1021A pin.

The interrupts are active low, but the GICv2 controller does not support active-low and falling-edge interrupts, so the only mode it can be configured in is rising-edge.

Quoted 7 lines

Hi Vladimir

So how does this work? The rising edge would occur after the interrupt handler has completed? What triggers the interrupt handler?

        Andrew

Quoted 7 lines

Hi Andrew,

I hope I am not terribly confused about this. I thought I am telling the interrupt controller to raise an IRQ as a result of the low-to-high transition of the electrical signal. Experimentation sure seems to agree with me. So the IRQ is generated immediately _after_ the PHY has left the line in open drain and it got pulled up to Vdd.

Hi Vladimir

                       t1                    t2

     ------------------\                     /----------------
                        \-------------------/

The interrupt output is active low. So it is high by default. At time t1 something happens, say the link is established. The interrupt becomes active, we have a failing edge. We want the interrupt controller to fire. Lets say it does. The interrupt handler runs, and clears the interrupt cause. This is at time t2. We then get a rising edge and the PHY releases the interrupt, and the level returns to high.

So how does this work if you have the interrupt controller triggering on a rising edge? The edge won't rise until the interrupt handler finishes its work.

Andrew

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

#3 lore raw
In reply to Vladimir Oltean
Quoted 17 lines · CET Vladimir Oltean
Quoted 8 lines · Andrew Lunn
Quoted 6 lines · Vladimir Oltean

On the LS1021A-TSN board, the 2 Atheros AR8031 PHYs for eth0 and eth1 have interrupt lines connected to the shared IRQ2_B LS1021A pin.

The interrupts are active low, but the GICv2 controller does not support active-low and falling-edge interrupts, so the only mode it can be configured in is rising-edge.

Quoted 7 lines

Hi Vladimir

So how does this work? The rising edge would occur after the interrupt handler has completed? What triggers the interrupt handler?

Andrew

Quoted 7 lines

Hi Andrew,

I hope I am not terribly confused about this. I thought I am telling the interrupt controller to raise an IRQ as a result of the low-to-high transition of the electrical signal. Experimentation sure seems to agree with me. So the IRQ is generated immediately _after_ the PHY has left the line in open drain and it got pulled up to Vdd.

It is correct GIC only supports raising edge and active-high. The IRQ[0:5] on ls1021a are a bit special though. They not directly connected to GIC, but there is an optional inverter, enabled by default. See RM for register SCFG_INTPCR. If left to default, those pins get actually active-high internally.
There was a patch 2 years ago to add support for this inverter: https://lore.kernel.org/patchwork/patch/860993/

Best regards,
Alexander

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

#4 lore raw
Quoted 27 lines · Alexander Stein
Quoted 17 lines · CET Vladimir Oltean
Quoted 8 lines · Andrew Lunn
Quoted 6 lines · Vladimir Oltean

On the LS1021A-TSN board, the 2 Atheros AR8031 PHYs for eth0 and eth1 have interrupt lines connected to the shared IRQ2_B LS1021A pin.

The interrupts are active low, but the GICv2 controller does not support active-low and falling-edge interrupts, so the only mode it can be configured in is rising-edge.

Quoted 7 lines

Hi Vladimir

So how does this work? The rising edge would occur after the interrupt handler has completed? What triggers the interrupt handler?

Andrew

Quoted 7 lines

Hi Andrew,

I hope I am not terribly confused about this. I thought I am telling the interrupt controller to raise an IRQ as a result of the low-to-high transition of the electrical signal. Experimentation sure seems to agree with me. So the IRQ is generated immediately _after_ the PHY has left the line in open drain and it got pulled up to Vdd.

Quoted 4 lines

It is correct GIC only supports raising edge and active-high. The IRQ[0:5] on ls1021a are a bit special though. They not directly connected to GIC, but there is an optional inverter, enabled by default.

Ah, O.K. So configuring for a rising edge is actually giving a falling edge. Which is why it works.

Actually supporting this correctly is going a cause some pain. I wonder how many DT files currently say rising/active high, when in fact falling/active low is actually being used? And when the IRQ controller really does support active low and falling, things brake?

Vladimir, since this is a shared interrupt, you really should use active low here. Maybe the first step is to get control of the inverter, and define a DT binding which is not going to break backwards compatibility. And then wire up this interrupt.

Andrew

#5 lore raw
Quoted 42 lines · Andrew Lunn
Quoted 27 lines · Alexander Stein
Quoted 17 lines · CET Vladimir Oltean
Quoted 8 lines · Andrew Lunn
Quoted 6 lines · Vladimir Oltean

On the LS1021A-TSN board, the 2 Atheros AR8031 PHYs for eth0 and eth1 have interrupt lines connected to the shared IRQ2_B LS1021A pin.

The interrupts are active low, but the GICv2 controller does not support active-low and falling-edge interrupts, so the only mode it can be configured in is rising-edge.

Quoted 7 lines

Hi Vladimir

So how does this work? The rising edge would occur after the interrupt handler has completed? What triggers the interrupt handler?

Andrew

Quoted 7 lines

Hi Andrew,

I hope I am not terribly confused about this. I thought I am telling the interrupt controller to raise an IRQ as a result of the low-to-high transition of the electrical signal. Experimentation sure seems to agree with me. So the IRQ is generated immediately _after_ the PHY has left the line in open drain and it got pulled up to Vdd.

Quoted 4 lines

It is correct GIC only supports raising edge and active-high. The IRQ[0:5] on ls1021a are a bit special though. They not directly connected to GIC, but there is an optional inverter, enabled by default.

Ah, O.K. So configuring for a rising edge is actually giving a falling edge. Which is why it works.

Actually supporting this correctly is going a cause some pain. I wonder how many DT files currently say rising/active high, when in fact falling/active low is actually being used? And when the IRQ controller really does support active low and falling, things brake?

Quoted 6 lines

Vladimir, since this is a shared interrupt, you really should use active low here. Maybe the first step is to get control of the inverter, and define a DT binding which is not going to break backwards compatibility. And then wire up this interrupt.

Andrew

Oh, ok, this is what you mean, thanks Alexander for the clarification. This sure escalated quickly and is going to keep me busy for a while.

-Vladimir

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

#6 lore raw
Quoted 49 lines · Vladimir Oltean
Quoted 47 lines · Andrew Lunn
Quoted 27 lines · Alexander Stein
Quoted 17 lines · CET Vladimir Oltean
Quoted 8 lines · Andrew Lunn
Quoted 6 lines · Vladimir Oltean

On the LS1021A-TSN board, the 2 Atheros AR8031 PHYs for eth0 and eth1 have interrupt lines connected to the shared IRQ2_B LS1021A pin.

The interrupts are active low, but the GICv2 controller does not support active-low and falling-edge interrupts, so the only mode it can be configured in is rising-edge.

Quoted 7 lines

Hi Vladimir

So how does this work? The rising edge would occur after the interrupt handler has completed? What triggers the interrupt handler?

Andrew

Quoted 7 lines

Hi Andrew,

I hope I am not terribly confused about this. I thought I am telling the interrupt controller to raise an IRQ as a result of the low-to-high transition of the electrical signal. Experimentation sure seems to agree with me. So the IRQ is generated immediately _after_ the PHY has left the line in open drain and it got pulled up to Vdd.

Quoted 4 lines

It is correct GIC only supports raising edge and active-high. The IRQ[0:5] on ls1021a are a bit special though. They not directly connected to GIC, but there is an optional inverter, enabled by default.

Ah, O.K. So configuring for a rising edge is actually giving a falling edge. Which is why it works.

Actually supporting this correctly is going a cause some pain. I wonder how many DT files currently say rising/active high, when in fact falling/active low is actually being used? And when the IRQ controller really does support active low and falling, things brake?

Vladimir, since this is a shared interrupt, you really should use active low here. Maybe the first step is to get control of the inverter, and define a DT binding which is not going to break backwards compatibility. And then wire up this interrupt.

Quoted 6 lines
Quoted 1 line

Andrew

Oh, ok, this is what you mean, thanks Alexander for the clarification. This sure escalated quickly and is going to keep me busy for a while.

-Vladimir

Sorry, I'm still a bit in shock, since this hit me in the face from nowhere, so I hadn't followed the entire history when I sent the above email.
It looks after all that Kurt and Rasmus have picked this up again and that the latest patch set is from 2 days ago, I'll take a look at that...
https://lwn.net/Articles/804103/

Thanks,
-Vladimir