Re: [PATCH] arm64: dts: mt7622: fix switch probe on bananapi-r64

From: Arınç ÜNAL
Date: Fri May 31 2024 - 02:27:22 EST


On 31/05/2024 09.18, Frank Wunderlich wrote:
Am 31. Mai 2024 08:12:06 MESZ schrieb "Arınç ÜNAL" <arinc.unal@xxxxxxxxxx>:
On 17/05/2024 09.27, Frank Wunderlich wrote:
Am 17. Mai 2024 04:17:47 MESZ schrieb "Arınç ÜNAL" <arinc.unal@xxxxxxxxxx>:
On 16/05/2024 23:48, Frank Wunderlich wrote:
From: Frank Wunderlich <frank-w@xxxxxxxxxxxxxxx>

After commit 868ff5f4944a
("net: dsa: mt7530-mdio: read PHY address of switch from device tree")
the mt7531 switch on Bananapi-R64 was not detected.

mt7530-mdio mdio-bus:00: reset timeout
mt7530-mdio mdio-bus:00: probe with driver mt7530-mdio failed with error -110

Fix this by adding phy address in devicetree.

Signed-off-by: Frank Wunderlich <frank-w@xxxxxxxxxxxxxxx>

I don't like the mention of the Linux kernel driver on the patch log. What
you're fixing is the incorrect description of the switch's PHY address on
the DTS file. Whether or not any driver from any project is actually
reading it from the DTS file is irrelevant to this patch. That said, I
already have a patch series I've been meaning to send the next version of
that already addresses this. Please wait for that.

Arınç

Hi arinc,

From my PoV it is a regression in next/6.10 because the driver change was merged (without "broadcast" fallback) and the dts patch [1] is not.

What is a broadcast fallback? 0x1f is just another PHY address.

Afaik 0x0 is some kind of broadcast address if real phy address is not known. The driver change seems not allow this 0x0 adress and forces devicetree to have the real address.

That's not true. 0x0 is just another PHY address. The driver change
actually allows reading 0x0 from the device tree and using that PHY address
to control the switch registers, instead of using 0x1f which was hardcoded
on the driver. But on the hardware, the switch actually listens on 0x1f.

Arınç