Ok, let's proceed :) The code runs well, dmesg looks good, ip addr shows me a link up, speed/duplex looks ok. But it does not transfer any data.
Debugging steps (A/B versions): - Check clocks with oscilloscope (10/100/1000) - Dump actual register settings - Trace Phy-Phy autonegotiation and ensure that our patch catches up the result Research topic: I will also look if in my test hardware uses distinct MDIO traces or in-band MDIO (MDIO via MII). Would inband MDIO need the autosense feature to talk to the phy before phy-phy auto negotiation is complete ? E.g. for triggering phy-phy autoneg ? And is this (inband mdio) maybe the reason why the may-phy autosense feature exists in the silicon ? Thanks, Roelof > Am 03.06.2020 um 17:59 schrieb Andrew Lunn <and...@lunn.ch>: > > On Wed, Jun 03, 2020 at 04:52:32PM +0200, Roelof Berg wrote: >> TEST REPORT: BROKEN PATCH >> >> Thanks to everyone for working on the fixed link feature of lan743x eth >> driver. >> > >> I received more test hardware today, and one piece of hardware >> (EVBlan7430) becomes incompatible by the patch. We need to roll back >> for now. Sorry. > > Hi Roelof > > We have a bit of time to fix this, before it becomes too critical. > So lets try to fix it. > > How did it break? > > Thanks > Andrew >
