[PATCH v4] dm: uclass: don't assign aliased seq numbers

[Michael Walle]

Hi Simon,

Am 2020-04-24 16:17, schrieb Michael Walle:
> Hi Simon,
> 
> Am 2020-04-20 01:38, schrieb Simon Glass:
> 
> [..snip..]
> 
>>> > uclass 31: eth
>>> > 0   * smsc95xx_eth @ 3db69ac0, seq 0, (req -1)
>>> 
>>> Shouldn't this be "req 0" if the ethernet alias is actually matched.
>>> Does u-boot actually supports matching usb nodes to devices? If not,
>>> shouldn't the alias be removed then?
>>> 
>>> That being said, it is still strange why the bootloader doesn't find
>>> ethernet-1 then. I've tried with my board, no native ethernet support
>>> and an usb network dongle which works as expected (well the dongle
>>> seems to have some issues to actually transfer frames).
>> 
>> It is a bit strange. Removing the alias does not fix it though.
> 
> Are you sure you removed the alias in the correct file? There are two,
> could you please double check if is not contained in the resulting
> device tree?
> 
> dtc -I dtb -O dts dts/dt.dtb
> 
> I just tested it on a rpi3b. and it works if i remove the alias.
> 
>> So far as I know U-Boot doesn't work with the alias, since there is no
>> driver for the "usb424,2514" compatible string.
> 
> So it is actually correct behaviour of my patch. ethernet1 doesn't work
> because there is no eth1addr. So I see three solutions:
> 
> (1) make the matching work
> (2) remove the alias
> (3) set eth1addr instead of ethaddr

Any news on this? Can I help somewhere? I'd go with (2).

-michael