Re: [fedora-arm] Banana Pi-R1 - kernel 5.6.0 and later broken - b43 DSA

[Florian Fainelli]

On 6/25/2020 4:23 PM, Peter Robinson wrote:
>> I'm having troubles with the Banana Pi-R1 router with newer kernels. No
>> config changes, config works well since a lot of lernel updates ...
>> Banana Pi-R1 is configured via systemd-networkd and uses the DSA
>> (Distributed Switch Architecture) with b53 switch. No visible difference
>> in interfaces, vlan config, bridge config, etc. Looks like actual
>> configuration in the switch in the hardware is broken.
>>
>> # OK: Last good known version (booting that version is also ok)
>> Linux bpi 5.5.18-200.fc31.armv7hl #1 SMP Fri Apr 17 17:25:00 UTC 2020
>> armv7l armv7l armv7l GNU/Linux
>>
>> # NOK: no network
>> Linux bpi 5.6.8-200.fc31.armv7hl #1 SMP Wed Apr 29 19:05:06 UTC 2020
>> armv7l armv7l armv7l GNU/Linux
>>
>> # NOK: no network
>> Linux bpi 5.6.0-300.fc32.armv7hl #1 SMP Mon Mar 30 16:37:50 UTC 2020
>> armv7l armv7l armv7l GNU/Linux
>>
>> # NOK: no network
>> Linux bpi 5.6.19-200.fc31.armv7hl #1 SMP Wed Jun 17 17:10:22 UTC 2020
>> armv7l armv7l armv7l GNU/Linux
>>
>> # NOK: no network
>> Linux bpi 5.7.4-200.fc32.armv7hl #1 SMP Fri Jun 19 00:52:22 UTC 2020
>> armv7l armv7l armv7l GNU/Linux
>>
>>
>> Saw that there were a lot of changes in the near past in the b53 driver:
>>
>> https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/log/drivers/net/dsa/b53?h=v5.8-rc1+
> 
> So from your notes above it looks like something regressed between the
> 5.5.x series and the 5.6.0 release? If so I'm not sure what the
> context of the changed in 5.8 are here. If there's more changes to the
> driver in 5.8 dev cycle have you tried one of the Fedora rawhide 5.8
> rc1 or rc2 kernel builds to see if that resolves it?

See my response to Gerhard, there are changes expected to how the
network configuration should be done after:

https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=8fab459e69abfd04a66d76423d18ba853fced4ab

> 
> Failing that looking for the change in the 5.6 cycle would likely be
> the better bet so we can work out what broke it and report it to the
> upstream driver maintainer. Have you searched to see if others are
> seeing a similar issue or have reported a similar issue upstream?
> 

Gerhard is doing exactly this with his email.
-- 
Florian