[fedora-arm] Each boot on my Cubietruck, and different MACaddr

[Robert Moskowitz]

Just an update.  I am getting a Cubietruck ready from production as a 
mailserver.

Everytime it boots it is getting a different MACaddr and thus the 
interface name advances one (eth1, eth2, eth3, etc.).  my ifcfg-eth0 
content IS being applied to this interface, so it is just that once I 
get things stable where I will not be booting a couple times a day, I 
can clean up the rules such that it will come up as eth1.

Going to have to run this way for a long time, as I don't think that 
Centos7-arm will be any better...

On 09/09/2014 07:12 PM, Hans de Goede wrote:
> Hi,
>
> On 09/09/2014 10:44 PM, Robert Moskowitz wrote:
>> OK.  back doing some testing.  So far all with F19.
>>
>> On 09/09/2014 07:02 AM, Hans de Goede wrote:
>>> Hi,
>>>
>>> On 09/09/2014 12:28 PM, Robert Moskowitz wrote:
>>>> On 09/09/2014 02:41 AM, Hans de Goede wrote:
>>>>> Hi,
>>>>>
>>>>> On 09/09/2014 07:40 AM, Marcin Juszkiewicz wrote:
>>>>>> W dniu 09.09.2014 o 04:51, Robert Moskowitz pisze:
>>>>>>> Read Hans de Goede's post on the F19 and the Allwinner (Cubie) back on
>>>>>>> 12/26/2013.  He supposedly uses the SID for a consistant local scope
>>>>>>> MACaddr.  I Do get that on my Cubieboard2, but not on my Cubietruck.
>>>>>> Hans also told that there was a bunch of Cubie* with pre-production cpus
>>>>>> which lacked serial number.
>>>>> AFAIK only cubieboard2 and olinuxino-a20-micro suffer from this, the
>>>>> cubietruck is fine. On machines with an all 0 sid this indeed does not
>>>>> work.
>>>> My SIX Cubieboard2 all are fine with a consistant MAC address.
>>> Good, then later runs of the cubieboard2 have gotten a proper SID, that is good.
>>>
>>>> My one Cubietruck keeps coming up with the unique MACaddr.  Exact opposite of what you are reporting.
>>> Which version of Fedora are you running on your Cubietruck ?
>>>
>>> With the respin images the kernel takes care of getting the MAC from the SID,
>>> with the official F-21 images, u-boot needs to do this, and the u-boot included
>>> is (not yet) new enough.
>>>
>>>> Perhaps there is a rev difference in the shipped nand?  In all cases, I never did anything with the nand image, but went right to the SDcard image.  I just don't get how to do the nand poking to get something working from it.
>>> The SID is in the SoC, it has nothing to do with the nand.
>>>
>>>> Is there any test I can run on my CT?
>>> You can dump the sid by hooking up a serial console, and then pressing a
>>> key to interrupt u-boot, then do:
>>>
>>> md 1c23800 4
>>>
>>> To get the SID contents, see here for example SID-s :
>>>
>>> http://linux-sunxi.org/SID_Register_Guide
>>>
>>> If this is non-0 on your cubietruck, then you should get a consistent MAC
>>> with the respin images,
>> Well, I am not.  For the Cubietruck.  Here is the SID fromthe CT and different MACaddr from a series of boots from the F19 image:
>>
>> md 1c23800 4
>>
>> 01c23800: 165166c1 80485172 49514848 0881e2d7    .fQ.rQH.HHQI....
>>
>> ea:99:af:e5:2f:5b
>>
>> ea:22:02:c6:83:c8
>>
>> fe:1d:9a:25:3b:48
>>
>> Then I switch over to a Cubieboard2, also with F19:
>>
>> md 1c23800 4
>> 01c23800: 165166c4 80485072 56484848 0382c153    .fQ.rPH.HHHVS...
>>
>> 02:c4:03:82:c1:53
>>
>>
>> I notice with the C2, the MAC is the last word of the SID.  But for the CT, I cannot figure it out at all.
> Ah right, I remember now the kernel code to use the SID for the MAC in F-19 is
> part of the emac driver, and the A20 uses the gmac driver, that is why it is
> not working on the cubietruck with F-19.
>
>> I have a bit of other testing to do (VLAN setup), then I will switch to F21 and see if I can get the current uboot for the CT working.
> Yes F-21 + latest u-boot should work.
>
> Regards,
>
> Hans
>