Skip to contentSkip to conversation

phy: phylink: Fix CuSFP issue in phylink

PATCH In review Bjarni Jonasson 4 participants 9 messages Nov 10, 2020 – Nov 17, 2020

Subsystems (4)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 phy: phylink: Fix CuSFP issue in phylink

There is an issue with the current phylink driver and CuSFPs which results in a callback to the phylink validate function without any advertisement capabilities. The workaround (in this changeset) is to assign capabilities if a 1000baseT SFP is identified.

---
 drivers/net/phy/phylink.c | 8 ++++++++
 1 file changed, 8 insertions(+)
drivers/net/phy/phylink.c +8 -0
index 32f4e8ec96cf..76e25f7f6934 100644
@@ -2196,6 +2196,14 @@ static int phylink_sfp_connect_phy(void *upstream, struct phy_device *phy)
 		mode = MLO_AN_INBAND;
 
 	/* Do the initial configuration */
+	if (phylink_test(pl->sfp_support, 1000baseT_Full)) {
+		pr_info("%s:%d: adding 1000baseT to PHY\n", __func__, __LINE__);
+		phylink_set(phy->supported, 1000baseT_Half);
+		phylink_set(phy->supported, 1000baseT_Full);
+		phylink_set(phy->advertising, 1000baseT_Half);
+		phylink_set(phy->advertising, 1000baseT_Full);
+	}
+
 	ret = phylink_sfp_config(pl, mode, phy->supported, phy->advertising);
 	if (ret < 0)
 		return ret;
Signature · 1 line
2.17.1
#1 lore raw
Quoted 4 lines · Bjarni Jonasson

There is an issue with the current phylink driver and CuSFPs which results in a callback to the phylink validate function without any advertisement capabilities. The workaround (in this changeset) is to assign capabilities if a 1000baseT SFP is identified.

How does this happen? Which PHY is being used?

Signature · 2 lines
RMK's Patch system: https://www.armlinux.org.uk/developer/patches/
FTTP is here! 40Mbps down 10Mbps up. Decent connectivity at last!
#2 lore raw
Quoted 7 lines · Russell King - ARM Linux admin
Quoted 4 lines · Bjarni Jonasson

There is an issue with the current phylink driver and CuSFPs which results in a callback to the phylink validate function without any advertisement capabilities. The workaround (in this changeset) is to assign capabilities if a 1000baseT SFP is identified.

How does this happen? Which PHY is being used?

This occurs just by plugging in the CuSFP.
None of the CuSFPs we have tested are working.
This is a dump from 3 different CuSFPs, phy regs 0-3:
FS SFP: 01:40:79:49
HP SFP: 01:40:01:49
Marvel SFP: 01:40:01:49
This was working before the delayed mac config was implemented (in dec 2019).

--
Bjarni Jonasson, Microchip

#3 lore raw
Quoted 11 lines · Bjarni Jonasson
Quoted 7 lines · Russell King - ARM Linux admin
Quoted 4 lines · Bjarni Jonasson

There is an issue with the current phylink driver and CuSFPs which results in a callback to the phylink validate function without any advertisement capabilities. The workaround (in this changeset) is to assign capabilities if a 1000baseT SFP is identified.

How does this happen? Which PHY is being used?

Quoted 8 lines

This occurs just by plugging in the CuSFP.
None of the CuSFPs we have tested are working.
This is a dump from 3 different CuSFPs, phy regs 0-3:
FS SFP: 01:40:79:49
HP SFP: 01:40:01:49
Marvel SFP: 01:40:01:49
This was working before the delayed mac config was implemented (in dec 2019).

You're dumping PHY registers 0 and 1 there, not 0 through 3, which the values confirm. I don't recognise the format either. PHY registers are always 16-bit.

Signature · 2 lines
RMK's Patch system: https://www.armlinux.org.uk/developer/patches/
FTTP is here! 40Mbps down 10Mbps up. Decent connectivity at last!
#4 lore raw
Quoted 21 lines · Russell King - ARM Linux admin
Quoted 11 lines · Bjarni Jonasson
Quoted 7 lines · Russell King - ARM Linux admin
Quoted 4 lines · Bjarni Jonasson

There is an issue with the current phylink driver and CuSFPs which results in a callback to the phylink validate function without any advertisement capabilities. The workaround (in this changeset) is to assign capabilities if a 1000baseT SFP is identified.

How does this happen? Which PHY is being used?

Quoted 8 lines

This occurs just by plugging in the CuSFP.
None of the CuSFPs we have tested are working.
This is a dump from 3 different CuSFPs, phy regs 0-3:
FS SFP: 01:40:79:49
HP SFP: 01:40:01:49
Marvel SFP: 01:40:01:49
This was working before the delayed mac config was implemented (in dec 2019).

Quoted 3 lines

You're dumping PHY registers 0 and 1 there, not 0 through 3, which the values confirm. I don't recognise the format either. PHY registers are always 16-bit.

Sorry about that. Here is it again:
Marvell SFP : 0x0140 0x0149 0x0141 0x0cc1
FS SFP : 0x1140 0x7949 0x0141 0x0cc2
Cisco SFP : 0x0140 0x0149 0x0141 0x0cc1
I.e. its seems to be a Marvell phy (0x0141) in all cases. And this occurs when phylink_start() is called. --
Bjarni Jonasson, Microchip

#5 lore raw
Quoted 27 lines · Bjarni Jonasson
Quoted 21 lines · Russell King - ARM Linux admin
Quoted 11 lines · Bjarni Jonasson
Quoted 7 lines · Russell King - ARM Linux admin
Quoted 4 lines · Bjarni Jonasson

There is an issue with the current phylink driver and CuSFPs which results in a callback to the phylink validate function without any advertisement capabilities. The workaround (in this changeset) is to assign capabilities if a 1000baseT SFP is identified.

How does this happen? Which PHY is being used?

Quoted 8 lines

This occurs just by plugging in the CuSFP.
None of the CuSFPs we have tested are working.
This is a dump from 3 different CuSFPs, phy regs 0-3:
FS SFP: 01:40:79:49
HP SFP: 01:40:01:49
Marvel SFP: 01:40:01:49
This was working before the delayed mac config was implemented (in dec 2019).

Quoted 3 lines

You're dumping PHY registers 0 and 1 there, not 0 through 3, which the values confirm. I don't recognise the format either. PHY registers are always 16-bit.

Quoted 6 lines

Sorry about that. Here is it again:
Marvell SFP : 0x0140 0x0149 0x0141 0x0cc1
FS SFP : 0x1140 0x7949 0x0141 0x0cc2
Cisco SFP : 0x0140 0x0149 0x0141 0x0cc1
I.e. its seems to be a Marvell phy (0x0141) in all cases. And this occurs when phylink_start() is called.

So they're all 88E1111 devices, which is the most common PHY for CuSFPs.

Do you have the Marvell PHY driver either built-in or available as a module? I suspect the problem is you don't. You will need the Marvell PHY driver to correctly drive the PHY, you can't rely on the fallback driver for SFPs.

Signature · 2 lines
RMK's Patch system: https://www.armlinux.org.uk/developer/patches/
FTTP is here! 40Mbps down 10Mbps up. Decent connectivity at last!
#6 lore raw
Quoted 35 lines · Russell King - ARM Linux admin
Quoted 27 lines · Bjarni Jonasson
Quoted 21 lines · Russell King - ARM Linux admin
Quoted 11 lines · Bjarni Jonasson
Quoted 7 lines · Russell King - ARM Linux admin
Quoted 4 lines · Bjarni Jonasson

There is an issue with the current phylink driver and CuSFPs which results in a callback to the phylink validate function without any advertisement capabilities. The workaround (in this changeset) is to assign capabilities if a 1000baseT SFP is identified.

How does this happen? Which PHY is being used?

Quoted 8 lines

This occurs just by plugging in the CuSFP.
None of the CuSFPs we have tested are working.
This is a dump from 3 different CuSFPs, phy regs 0-3:
FS SFP: 01:40:79:49
HP SFP: 01:40:01:49
Marvel SFP: 01:40:01:49
This was working before the delayed mac config was implemented (in dec 2019).

Quoted 3 lines

You're dumping PHY registers 0 and 1 there, not 0 through 3, which the values confirm. I don't recognise the format either. PHY registers are always 16-bit.

Quoted 6 lines

Sorry about that. Here is it again:
Marvell SFP : 0x0140 0x0149 0x0141 0x0cc1
FS SFP : 0x1140 0x7949 0x0141 0x0cc2
Cisco SFP : 0x0140 0x0149 0x0141 0x0cc1
I.e. its seems to be a Marvell phy (0x0141) in all cases. And this occurs when phylink_start() is called.

Quoted 7 lines

So they're all 88E1111 devices, which is the most common PHY for CuSFPs.

Do you have the Marvell PHY driver either built-in or available as a module? I suspect the problem is you don't. You will need the Marvell PHY driver to correctly drive the PHY, you can't rely on the fallback driver for SFPs.

Correct. I was using the generic driver and that does clearly not work. After including the Marvell driver the callback to the validate function happens as expected. Thanks for the support.
--
Bjarni Jonasson Microchip

maintainer
#7 lore raw
Quoted 7 lines
Quoted 4 lines

Do you have the Marvell PHY driver either built-in or available as a module? I suspect the problem is you don't. You will need the Marvell PHY driver to correctly drive the PHY, you can't rely on the fallback driver for SFPs.

Correct. I was using the generic driver and that does clearly not work. After including the Marvell driver the callback to the validate function happens as expected. Thanks for the support.

Hi Russell

Maybe we should have MDIO_I2C driver select the Marvell PHY driver?

Andrew

#8 lore raw
Quoted 11 lines · Andrew Lunn
Quoted 7 lines
Quoted 4 lines

Do you have the Marvell PHY driver either built-in or available as a module? I suspect the problem is you don't. You will need the Marvell PHY driver to correctly drive the PHY, you can't rely on the fallback driver for SFPs.

Correct. I was using the generic driver and that does clearly not work. After including the Marvell driver the callback to the validate function happens as expected. Thanks for the support.

Hi Russell

Maybe we should have MDIO_I2C driver select the Marvell PHY driver?

It was suggested a while ago that MARVELL_PHY follow the SFP configuration symbol and that we would warn when a CuSFP module was used with the Generic PHY driver:

https://www.mail-archive.com/[email protected]/msg253839.html

Eventually we did not make progress towards creating a list of modules that would require a specialized PHY driver, maybe we can start now?

Signature · 1 line
Florian