Skip to contentSkip to conversation

net: phy: realtek: fix wake-on-lan support

RFC PATCH In review Russell King (Oracle) 6 participants 9 messages Jul 30, 2025

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

RFC PATCH 1/1 net: phy: realtek: fix wake-on-lan support

Implement Wake-on-Lan correctly. The existing implementation has multiple issues:

1. It assumes that Wake-on-Lan can always be used, whether or not the interrupt is wired, and whether or not the interrupt is capable of waking the system. This breaks the ability for MAC drivers to detect whether the PHY WoL is functional.
2. switching the interrupt pin in the .set_wol() method to PMEB mode immediately silences link-state interrupts, which breaks phylib when interrupts are being used rather than polling mode.
3. the code claiming to "reset WOL status" was doing nothing of the sort. Bit 15 in page 0xd8a register 17 controls WoL reset, and needs to be pulsed low to reset the WoL state. This bit was always written as '1', resulting in no reset.
4. not resetting WoL state results in the PMEB pin remaining asserted, which in turn leads to an interrupt storm. Only resetting the WoL state in .set_wol() is not sufficient.
5. PMEB mode does not allow software detection of the wake-up event as there is no status bit to indicate we received the WoL packet.
6. across reboots of at least the Jetson Xavier NX system, the WoL configuration is preserved.

Fix all of these issues by essentially rewriting the support. We:
1. clear the WoL event enable register at probe time.
2. detect whether we can support wake-up by having a valid interrupt, and the "wakeup-source" property in DT. If we can, then we mark the MDIO device as wakeup capable, and associate the interrupt with the wakeup source.
3. arrange for the get_wol() and set_wol() implementations to handle the case where the MDIO device has not been marked as wakeup capable (thereby returning no WoL support, and refusing to enable WoL support.)
4. avoid switching to PMEB mode, instead using INTB mode with the interrupt enable, reconfiguring the interrupt enables at suspend time, and restoring their original state at resume time (we track the state of the interrupt enable register in .config_intr() register.)
5. move WoL reset from .set_wol() to the suspend function to ensure that WoL state is cleared prior to suspend. This is necessary after the PME interrupt has been enabled as a second WoL packet will not re-raise a previously cleared PME interrupt.
6. when a PME interrupt (for wakeup) is asserted, pass this to the PM wakeup so it knows which device woke the system.

---
I've sort-of tested this on the Jetson Xavier NX platform, but it's been difficult because it appears that the whole interrupt/wakeup stuff for the SoC is foobar in mainline. One gets the choice of specifying the GPIO interrupt in DT and have working normal interrupt or the power management controller interrupt for the same line and having wakeup functional. You can't have both together.

I'm not sure whether this change should target the net or net-next tree; what we have currently in 6.16 is totally and utterly broken, so arguably this is a fix - but it's not a regression because 6.16 is the first kernel that WoL "support" for RTL8211F is in. This is also a large change.

However, I can't see that it was tested, given all the problems identified above. As a result, I've taken the decision in this patch to not worry about breaking anyone's existing setup.

So, I have no problem with requiring "wakeup-source" to be added to DT for rtl8211f PHYs that are to support wake-up, meaning that they are properly wired to support WoL. We can't tell just from having an interrupt - not all interrupts on all devices may be wake-up capable.

 drivers/net/phy/realtek/realtek_main.c | 174 ++++++++++++++++++++-----
 1 file changed, 142 insertions(+), 32 deletions(-)
drivers/net/phy/realtek/realtek_main.c +142 -32
index dd0d675149ad..520fdeb88c32 100644
@@ -10,6 +10,7 @@
 #include <linux/bitops.h>
 #include <linux/of.h>
 #include <linux/phy.h>
+#include <linux/pm_wakeirq.h>
 #include <linux/netdevice.h>
 #include <linux/module.h>
 #include <linux/delay.h>
@@ -31,6 +32,7 @@
 #define RTL821x_INER				0x12
 #define RTL8211B_INER_INIT			0x6400
 #define RTL8211E_INER_LINK_STATUS		BIT(10)
+#define RTL8211F_INER_PME			BIT(7)
 #define RTL8211F_INER_LINK_STATUS		BIT(4)
 
 #define RTL821x_INSR				0x13
@@ -96,17 +98,13 @@
 #define RTL8211F_RXCR				0x15
 #define RTL8211F_RX_DELAY			BIT(3)
 
-/* RTL8211F WOL interrupt configuration */
-#define RTL8211F_INTBCR_PAGE			0xd40
-#define RTL8211F_INTBCR				0x16
-#define RTL8211F_INTBCR_INTB_PMEB		BIT(5)
-
 /* RTL8211F WOL settings */
-#define RTL8211F_WOL_SETTINGS_PAGE		0xd8a
+#define RTL8211F_WOL_PAGE		0xd8a
 #define RTL8211F_WOL_SETTINGS_EVENTS		16
 #define RTL8211F_WOL_EVENT_MAGIC		BIT(12)
-#define RTL8211F_WOL_SETTINGS_STATUS		17
-#define RTL8211F_WOL_STATUS_RESET		(BIT(15) | 0x1fff)
+#define RTL8211F_WOL_RST_RMSQ		17
+#define RTL8211F_WOL_RG_RSTB			BIT(15)
+#define RTL8211F_WOL_RMSQ			0x1fff
 
 /* RTL8211F Unique phyiscal and multicast address (WOL) */
 #define RTL8211F_PHYSICAL_ADDR_PAGE		0xd8c
@@ -172,7 +170,8 @@ struct rtl821x_priv {
 	u16 phycr2;
 	bool has_phycr2;
 	struct clk *clk;
-	u32 saved_wolopts;
+	/* rtl8211f */
+	u16 iner;
 };
 
 static int rtl821x_read_page(struct phy_device *phydev)
@@ -255,6 +254,35 @@ static int rtl821x_probe(struct phy_device *phydev)
 	return 0;
 }
 
+static int rtl8211f_probe(struct phy_device *phydev)
+{
+	struct device *dev = &phydev->mdio.dev;
+	int ret;
+
+	ret = rtl821x_probe(phydev);
+	if (ret < 0)
+		return ret;
+
+	/* Disable all PME events */
+	ret = phy_write_paged(phydev, RTL8211F_WOL_PAGE,
+			      RTL8211F_WOL_SETTINGS_EVENTS, 0);
+	if (ret < 0)
+		return ret;
+
+	/* Mark this PHY as wakeup capable and register the interrupt as a
+	 * wakeup IRQ if the PHY is marked as a wakeup source in firmware,
+	 * and the interrupt is valid. Save the INTB/PMEB pin coniguration,
+	 * so it can be restored when resuming.
+	 */
+	if (device_property_read_bool(dev, "wakeup-source") &&
+	    phy_interrupt_is_valid(phydev)) {
+		device_set_wakeup_capable(dev, true);
+		devm_pm_set_wake_irq(dev, phydev->irq);
+	}
+
+	return ret;
+}
+
 static int rtl8201_ack_interrupt(struct phy_device *phydev)
 {
 	int err;
@@ -352,6 +380,7 @@ static int rtl8211e_config_intr(struct phy_device *phydev)
 
 static int rtl8211f_config_intr(struct phy_device *phydev)
 {
+	struct rtl821x_priv *priv = phydev->priv;
 	u16 val;
 	int err;
 
@@ -362,8 +391,10 @@ static int rtl8211f_config_intr(struct phy_device *phydev)
 
 		val = RTL8211F_INER_LINK_STATUS;
 		err = phy_write_paged(phydev, 0xa42, RTL821x_INER, val);
+		if (err == 0)
+			priv->iner = val;
 	} else {
-		val = 0;
+		priv->iner = val = 0;
 		err = phy_write_paged(phydev, 0xa42, RTL821x_INER, val);
 		if (err)
 			return err;
@@ -426,21 +457,34 @@ static irqreturn_t rtl8211f_handle_interrupt(struct phy_device *phydev)
 		return IRQ_NONE;
 	}
 
-	if (!(irq_status & RTL8211F_INER_LINK_STATUS))
-		return IRQ_NONE;
+	if (irq_status & RTL8211F_INER_LINK_STATUS) {
+		phy_trigger_machine(phydev);
+		return IRQ_HANDLED;
+	}
 
-	phy_trigger_machine(phydev);
+	if (irq_status & RTL8211F_INER_PME) {
+		pm_wakeup_event(&phydev->mdio.dev, 0);
+		return IRQ_HANDLED;
+	}
 
-	return IRQ_HANDLED;
+	return IRQ_NONE;
 }
 
 static void rtl8211f_get_wol(struct phy_device *dev, struct ethtool_wolinfo *wol)
 {
 	int wol_events;
 
+	/* If the PHY is not capable of waking the system, then WoL can not
+	 * be supported.
+	 */
+	if (!device_can_wakeup(&dev->mdio.dev)) {
+		wol->supported = 0;
+		return;
+	}
+
 	wol->supported = WAKE_MAGIC;
 
-	wol_events = phy_read_paged(dev, RTL8211F_WOL_SETTINGS_PAGE, RTL8211F_WOL_SETTINGS_EVENTS);
+	wol_events = phy_read_paged(dev, RTL8211F_WOL_PAGE, RTL8211F_WOL_SETTINGS_EVENTS);
 	if (wol_events < 0)
 		return;
 
@@ -453,6 +497,9 @@ static int rtl8211f_set_wol(struct phy_device *dev, struct ethtool_wolinfo *wol)
 	const u8 *mac_addr = dev->attached_dev->dev_addr;
 	int oldpage;
 
+	if (!device_can_wakeup(&dev->mdio.dev))
+		return -EOPNOTSUPP;
+
 	oldpage = phy_save_page(dev);
 	if (oldpage < 0)
 		goto err;
@@ -464,25 +511,23 @@ static int rtl8211f_set_wol(struct phy_device *dev, struct ethtool_wolinfo *wol)
 		__phy_write(dev, RTL8211F_PHYSICAL_ADDR_WORD1, mac_addr[3] << 8 | (mac_addr[2]));
 		__phy_write(dev, RTL8211F_PHYSICAL_ADDR_WORD2, mac_addr[5] << 8 | (mac_addr[4]));
 
-		/* Enable magic packet matching and reset WOL status */
-		rtl821x_write_page(dev, RTL8211F_WOL_SETTINGS_PAGE);
+		/* Enable magic packet matching */
+		rtl821x_write_page(dev, RTL8211F_WOL_PAGE);
 		__phy_write(dev, RTL8211F_WOL_SETTINGS_EVENTS, RTL8211F_WOL_EVENT_MAGIC);
-		__phy_write(dev, RTL8211F_WOL_SETTINGS_STATUS, RTL8211F_WOL_STATUS_RESET);
-
-		/* Enable the WOL interrupt */
-		rtl821x_write_page(dev, RTL8211F_INTBCR_PAGE);
-		__phy_set_bits(dev, RTL8211F_INTBCR, RTL8211F_INTBCR_INTB_PMEB);
+		/* Set the maximum packet size, and assert WoL reset */
+		__phy_write(dev, RTL8211F_WOL_RST_RMSQ, RTL8211F_WOL_RMSQ);
 	} else {
-		/* Disable the WOL interrupt */
-		rtl821x_write_page(dev, RTL8211F_INTBCR_PAGE);
-		__phy_clear_bits(dev, RTL8211F_INTBCR, RTL8211F_INTBCR_INTB_PMEB);
-
-		/* Disable magic packet matching and reset WOL status */
-		rtl821x_write_page(dev, RTL8211F_WOL_SETTINGS_PAGE);
+		/* Disable magic packet matching */
+		rtl821x_write_page(dev, RTL8211F_WOL_PAGE);
 		__phy_write(dev, RTL8211F_WOL_SETTINGS_EVENTS, 0);
-		__phy_write(dev, RTL8211F_WOL_SETTINGS_STATUS, RTL8211F_WOL_STATUS_RESET);
+
+		/* Place WoL in reset */
+		__phy_clear_bits(dev, RTL8211F_WOL_RST_RMSQ,
+				 RTL8211F_WOL_RG_RSTB);
 	}
 
+	device_set_wakeup_enable(&dev->mdio.dev, !!(wol->wolopts & WAKE_MAGIC));
+
 err:
 	return phy_restore_page(dev, oldpage, 0);
 }
@@ -628,6 +673,52 @@ static int rtl821x_suspend(struct phy_device *phydev)
 	return ret;
 }
 
+static int rtl8211f_suspend(struct phy_device *phydev)
+{
+	u16 wol_rst;
+	int ret;
+
+	ret = rtl821x_suspend(phydev);
+	if (ret < 0)
+		return ret;
+
+	/* If a PME event is enabled, then configure the interrupt for
+	 * PME events only, disabling link interrupt. We avoid switching
+	 * to PMEB mode as we don't have a status bit for that.
+	 */
+	if (device_may_wakeup(&phydev->mdio.dev)) {
+		ret = phy_write_paged(phydev, 0xa42, RTL821x_INER,
+				      RTL8211F_INER_PME);
+		if (ret < 0)
+			goto err;
+
+		/* Read the INSR to clear any pending interrupt */
+		phy_read_paged(phydev, RTL8211F_INSR_PAGE, RTL8211F_INSR);
+
+		/* Reset the WoL to ensure that an event is picked up.
+		 * Unless we do this, even if we receive another packet,
+		 * we may not have a PME interrupt raised.
+		 */
+		ret = phy_read_paged(phydev, RTL8211F_WOL_PAGE,
+				     RTL8211F_WOL_RST_RMSQ);
+		if (ret < 0)
+			goto err;
+
+		wol_rst = ret & ~RTL8211F_WOL_RG_RSTB;
+		ret = phy_write_paged(phydev, RTL8211F_WOL_PAGE,
+				      RTL8211F_WOL_RST_RMSQ, wol_rst);
+		if (ret < 0)
+			goto err;
+
+		wol_rst |= RTL8211F_WOL_RG_RSTB;
+		ret = phy_write_paged(phydev, RTL8211F_WOL_PAGE,
+				      RTL8211F_WOL_RST_RMSQ, wol_rst);
+	}
+
+err:
+	return ret;
+}
+
 static int rtl821x_resume(struct phy_device *phydev)
 {
 	struct rtl821x_priv *priv = phydev->priv;
@@ -645,6 +736,25 @@ static int rtl821x_resume(struct phy_device *phydev)
 	return 0;
 }
 
+static int rtl8211f_resume(struct phy_device *phydev)
+{
+	struct rtl821x_priv *priv = phydev->priv;
+	int ret;
+
+	ret = rtl821x_resume(phydev);
+	if (ret < 0)
+		return ret;
+
+	/* If the device was programmed for a PME event, restore the interrupt
+	 * enable and interrupt pin configuration state so phylib can receive
+	 * link state interrupts.
+	 */
+	if (device_may_wakeup(&phydev->mdio.dev))
+		ret = phy_write_paged(phydev, 0xa42, RTL821x_INER, priv->iner);
+
+	return ret;
+}
+
 static int rtl8211x_led_hw_is_supported(struct phy_device *phydev, u8 index,
 					unsigned long rules)
 {
@@ -1612,15 +1722,15 @@ static struct phy_driver realtek_drvs[] = {
 	}, {
 		PHY_ID_MATCH_EXACT(0x001cc916),
 		.name		= "RTL8211F Gigabit Ethernet",
-		.probe		= rtl821x_probe,
+		.probe		= rtl8211f_probe,
 		.config_init	= &rtl8211f_config_init,
 		.read_status	= rtlgen_read_status,
 		.config_intr	= &rtl8211f_config_intr,
 		.handle_interrupt = rtl8211f_handle_interrupt,
 		.set_wol	= rtl8211f_set_wol,
 		.get_wol	= rtl8211f_get_wol,
-		.suspend	= rtl821x_suspend,
-		.resume		= rtl821x_resume,
+		.suspend	= rtl8211f_suspend,
+		.resume		= rtl8211f_resume,
 		.read_page	= rtl821x_read_page,
 		.write_page	= rtl821x_write_page,
 		.flags		= PHY_ALWAYS_CALL_SUSPEND,
Signature · 1 line
2.30.2
maintainer
#1 lore raw
Quoted 4 lines

2. detect whether we can support wake-up by having a valid interrupt, and the "wakeup-source" property in DT. If we can, then we mark the MDIO device as wakeup capable, and associate the interrupt with the wakeup source.

We should document "wakeup-source" in ethernet-phy.yaml.

What are the different hardware architectures?

1) A single interrupt line from the PHY to the SoC, which does both link status and WoL.

2) The PHY has a dedicated WoL output pin, which is connected to an interrupt.

3) The PHY has a dedicated WoL output pin, which is connected directly to a PMIC. No software involved, the pin toggling turns the power back on.

For 1), i don't think 'wakeup-source' tells us anything useful. The driver just needs to check that interrupts are in use.

For 2) we should recommend that the wakeup interrupt is called "wakeup", following wakeup-source.txt, and the "wakeup-source" property is present.

For 3) its more magical, there is no interrupt properties involved, so we do need the "wakeup-source" to know that the pin is actually connected to something.

We need to differentiate between drivers newly getting WoL support, and existing drivers. We can be much more strict with new support.

Andrew

#2 lore raw
Quoted 15 lines · Andrew Lunn
Quoted 4 lines

2. detect whether we can support wake-up by having a valid interrupt, and the "wakeup-source" property in DT. If we can, then we mark the MDIO device as wakeup capable, and associate the interrupt with the wakeup source.

We should document "wakeup-source" in ethernet-phy.yaml.

What are the different hardware architectures?

1) A single interrupt line from the PHY to the SoC, which does both link status and WoL.

2) The PHY has a dedicated WoL output pin, which is connected to an interrupt.

Quoted 6 lines

3) The PHY has a dedicated WoL output pin, which is connected directly to a PMIC. No software involved, the pin toggling turns the power back on.

For 1), i don't think 'wakeup-source' tells us anything useful. The driver just needs to check that interrupts are in use.

Not all interrupts are capable of waking the system up, and there is no way for a PHY to know whether it's connected to an interrupt that has that ability.

As things currently stand with how Jetson Xavier NX DT describes the PHY's connection, it falls into this "it has an interrupt which can't wake the system" - so these cases really do exist in the real world.

So, we _need_ to have some way to differentiate these two cases, and I put the question to you - if not by "wakeup-source" then how do we determine whether a PHY, which itself is capable of signalling wakeup through its interrupt pin, can actually wake the system, and thus should expose WoL functionality?

Signature · 2 lines
RMK's Patch system: https://www.armlinux.org.uk/developer/patches/
FTTP is here! 80Mbps down 10Mbps up. Decent connectivity at last!
maintainer
#3 lore raw
Quoted 3 lines

Not all interrupts are capable of waking the system up, and there is no way for a PHY to know whether it's connected to an interrupt that has that ability.

I was wondering about that. And maybe that enable_irq_wake() returns -EOPNOTSUPP if it cannot actually wake the system? But there is no documentation about that.

So we cannot trust it, and "wakeup-source" is needed.

Andrew

#4 lore raw
Quoted 7 lines · Andrew Lunn
Quoted 3 lines

Not all interrupts are capable of waking the system up, and there is no way for a PHY to know whether it's connected to an interrupt that has that ability.

I was wondering about that. And maybe that enable_irq_wake() returns -EOPNOTSUPP if it cannot actually wake the system? But there is no documentation about that.

,,, and that means we can't use the wakeirq helpers.

Signature · 2 lines
RMK's Patch system: https://www.armlinux.org.uk/developer/patches/
FTTP is here! 80Mbps down 10Mbps up. Decent connectivity at last!
#5 lore raw
In reply to Andrew Lunn
Quoted 7 lines · Andrew Lunn
Quoted 4 lines

2. detect whether we can support wake-up by having a valid interrupt, and the "wakeup-source" property in DT. If we can, then we mark the MDIO device as wakeup capable, and associate the interrupt with the wakeup source.

We should document "wakeup-source" in ethernet-phy.yaml.

+1

Quoted 5 lines

What are the different hardware architectures?

1) A single interrupt line from the PHY to the SoC, which does both link status and WoL.

Quoted 7 lines

2) The PHY has a dedicated WoL output pin, which is connected to an interrupt.

3) The PHY has a dedicated WoL output pin, which is connected directly to a PMIC. No software involved, the pin toggling turns the power back on.

Just my 2 cents:

In some cases like the LAN8742, there are some flags that need to be cleared when waking up in order to be able to handle another WoL event. It can be done either in the suspend()/resume() or in an interrupt handler of the PHY. In 3) This suggests that the interrupt is somehow forwarded to the Linux kernel.

This is what I was ultimately trying to do in two steps with the TEE notifying the kernel of that interrupt.

Moreover, if a WoL event occurs when the system is not in a low-power mode, then the flags will never be cleared and another WoL event cannot be detected while the system is in a low-power mode.

Maybe we can argue that these flags can be cleared in suspend() and and resume(). But then, if there's no interrupt to be handled by the kernel, how do we know that we have woken up from a WoL event?

IMHO, I think 3) may optionally declare another interrupt as well for WoL events.

Eventually, 2) and 3) would have 1 interrupt(WoL) if PHY is in polling mode and 2 if not?

Please tell me if that makes any sense to you or if I missed something.

Quoted 15 lines

For 1), i don't think 'wakeup-source' tells us anything useful. The driver just needs to check that interrupts are in use.

For 2) we should recommend that the wakeup interrupt is called "wakeup", following wakeup-source.txt, and the "wakeup-source" property is present.

For 3) its more magical, there is no interrupt properties involved, so we do need the "wakeup-source" to know that the pin is actually connected to something.

We need to differentiate between drivers newly getting WoL support, and existing drivers. We can be much more strict with new support.

Andrew

maintainer
#6 lore raw
Quoted 20 lines
Quoted 4 lines

3) The PHY has a dedicated WoL output pin, which is connected directly to a PMIC. No software involved, the pin toggling turns the power back on.

Just my 2 cents:

In some cases like the LAN8742, there are some flags that need to be cleared when waking up in order to be able to handle another WoL event. It can be done either in the suspend()/resume() or in an interrupt handler of the PHY. In 3) This suggests that the interrupt is somehow forwarded to the Linux kernel.

This is what I was ultimately trying to do in two steps with the TEE notifying the kernel of that interrupt.

Moreover, if a WoL event occurs when the system is not in a low-power mode, then the flags will never be cleared and another WoL event cannot be detected while the system is in a low-power mode.

Quoted 6 lines

Maybe we can argue that these flags can be cleared in suspend() and and resume(). But then, if there's no interrupt to be handled by the kernel, how do we know that we have woken up from a WoL event?

IMHO, I think 3) may optionally declare another interrupt as well for WoL events.

The 3) i've seen is a Marvell based NAS box. It is a long time ago. But as far as i remember, the SoC had no idea why it woke up, it could not ask the PMIC why the power was turned on. So there was no ability to invoke an interrupt handler.

I've also seen X86/BIOS platforms which are similar. The BIOS swallows the interrupt, it never gets delivered to Linux once it is running. And the BIOS itself might poke PHY registers.

Quoted 2 lines

Eventually, 2) and 3) would have 1 interrupt(WoL) if PHY is in polling mode and 2 if not?

Nope. These NAS boxed often had 0 PHY interrupts, and polling was used. If you have a single ethernet port, you don't care it takes a while to know the link is dead. There is nothing you can do, you don't have a second interface to fall back to.

Andrew

#7 lore raw
Quoted 11 lines
Quoted 2 lines

IMHO, I think 3) may optionally declare another interrupt as well for WoL events.

The 3) i've seen is a Marvell based NAS box. It is a long time ago. But as far as i remember, the SoC had no idea why it woke up, it could not ask the PMIC why the power was turned on. So there was no ability to invoke an interrupt handler.

I've also seen X86/BIOS platforms which are similar. The BIOS swallows the interrupt, it never gets delivered to Linux once it is running. And the BIOS itself might poke PHY registers.

That's like our hardware looks like.

We have a TI J784S4 SoC which has an integrated 9 port switch. One internal, and 8 external ports.

It's similar to the device tree in arch/arm64/boot/dts/ti/k3-j784s4-evm-quad-port-eth-exp1.dtso.

The PHYs are directly connected to the PMIC of our board.
We don't have any clue **why** the PMIC wakes the board.
Could be a WoL event or somebody pressed the power button of our robot controller.

Regards
Daniel

Internal

#8 lore raw
In reply to Russell King (Oracle)
Quoted 44 lines · Russell King (Oracle)

Implement Wake-on-Lan correctly. The existing implementation has multiple issues:

1. It assumes that Wake-on-Lan can always be used, whether or not the interrupt is wired, and whether or not the interrupt is capable of waking the system. This breaks the ability for MAC drivers to detect whether the PHY WoL is functional.
2. switching the interrupt pin in the .set_wol() method to PMEB mode immediately silences link-state interrupts, which breaks phylib when interrupts are being used rather than polling mode.
3. the code claiming to "reset WOL status" was doing nothing of the sort. Bit 15 in page 0xd8a register 17 controls WoL reset, and needs to be pulsed low to reset the WoL state. This bit was always written as '1', resulting in no reset.
4. not resetting WoL state results in the PMEB pin remaining asserted, which in turn leads to an interrupt storm. Only resetting the WoL state in .set_wol() is not sufficient.
5. PMEB mode does not allow software detection of the wake-up event as there is no status bit to indicate we received the WoL packet.
6. across reboots of at least the Jetson Xavier NX system, the WoL configuration is preserved.

Fix all of these issues by essentially rewriting the support. We:
1. clear the WoL event enable register at probe time.
2. detect whether we can support wake-up by having a valid interrupt, and the "wakeup-source" property in DT. If we can, then we mark the MDIO device as wakeup capable, and associate the interrupt with the wakeup source.
3. arrange for the get_wol() and set_wol() implementations to handle the case where the MDIO device has not been marked as wakeup capable (thereby returning no WoL support, and refusing to enable WoL support.)
4. avoid switching to PMEB mode, instead using INTB mode with the interrupt enable, reconfiguring the interrupt enables at suspend time, and restoring their original state at resume time (we track the state of the interrupt enable register in .config_intr() register.)
5. move WoL reset from .set_wol() to the suspend function to ensure that WoL state is cleared prior to suspend. This is necessary after the PME interrupt has been enabled as a second WoL packet will not re-raise a previously cleared PME interrupt.
6. when a PME interrupt (for wakeup) is asserted, pass this to the PM wakeup so it knows which device woke the system.

Quoted 1 line

This looks much better and straightforward to me, thanks!

Quoted 14 lines

---
I've sort-of tested this on the Jetson Xavier NX platform, but it's been difficult because it appears that the whole interrupt/wakeup stuff for the SoC is foobar in mainline. One gets the choice of specifying the GPIO interrupt in DT and have working normal interrupt or the power management controller interrupt for the same line and having wakeup functional. You can't have both together.

I'm not sure whether this change should target the net or net-next tree; what we have currently in 6.16 is totally and utterly broken, so arguably this is a fix - but it's not a regression because 6.16 is the first kernel that WoL "support" for RTL8211F is in. This is also a large change.

Quoted 8 lines

However, I can't see that it was tested, given all the problems identified above. As a result, I've taken the decision in this patch to not worry about breaking anyone's existing setup.

So, I have no problem with requiring "wakeup-source" to be added to DT for rtl8211f PHYs that are to support wake-up, meaning that they are properly wired to support WoL. We can't tell just from having an interrupt - not all interrupts on all devices may be wake-up capable.

That seems entirely reasonable to me to expect.

Signature · 1 line
Florian