ClearFog GT 8K not initialising SFP-H10GB-CU1M transceiver on 5.4.150

[Russell King (Oracle)]

On Tue, Nov 16, 2021 at 02:11:54PM +0400, Jordan Vrtanoski wrote:
> Hi,
> 
> I am trying to build kernel 5.4.150 for SolidRun ClearFog GT 8K. The kernel is stable and I had not experiance any stability issue. The only issue I am facing is with enabling the support for the SFP+ module. 
> 
> I am using SFP-H10GB-CU1M direct coper cable module. The module works on earlier version of the kernel 5.1.x (verified on the same hardware), however once the device is booted under 5.4.150, the transceiver is not recognised and the interface remains down. There are no error messages reported by the kernel.

The transceiver is recognised, but for some reason the link isn't coming
up - and from the debug kernel messages, there's no information pointing
to why not. At a guess, mvpp2 is not reporting that it has link for some
reason.

I don't think it's an issue with the module or the SFP layer - since
that is going into "link up" state and the TX_DISABLE signal is being
deasserted (not that they mean much for a DA cable.) So, I'd be thinking
that the problem is with the comphy or mvpp2 drivers.

mvpp2 doesn't have much in the way of debugging messages to help with
this, but phylink should at least be printing "mac link" messages in
both 5.1.x and 5.4.x kernels - but it doesn't appear to be. That
suggests enabling dynamic debug is not sufficient to get those messages
out... I'm afraid I've never used dynamic debug so can't help with that.
Maybe someone who knows now netdev_dbg() interacts with dynamic debug
can help with that. Strangely, you are getting the netdev_dbg() from
within phylink_sfp_module_insert(), so I don't really understand why
you aren't getting anything else from phylink.

It all seems rather odd.

-- 
RMK's Patch system: https://www.armlinux.org.uk/developer/patches/
FTTP is here! 40Mbps down 10Mbps up. Decent connectivity at last!