[U-Boot] [PATCH] net: use random ethernet address if invalid and not zero

[Joe Hershberger]

On Mon, Nov 7, 2016 at 4:25 PM, James Chargin <jimccrown at gmail.com> wrote:
> Hi Joe,
>
> Thanks for your reply.
>
> On 11/07/2016 10:20 AM, Joe Hershberger wrote:
>>
>> Hi Jim,
>>
>> On Wed, Nov 2, 2016 at 8:21 AM, James Chargin <jimccrown at gmail.com> wrote:
>>>
>>> Hi,
>>>
>>> Regarding "invalid" Ethernet address.
>>>
>>> Is there a reliable way to set the default environment that will prevent
>>> Ethernet communications from being attempted.
>>>
>>> That is, when an Ethernet capable system is brand new and before an
>>> Ethernet
>>> MAC address has been assigned to that system during manufacturing with
>>> the
>>> "setenv ethaddr" command, how can Ethernet comms be disabled?
>>>
>>> This would be a fail-safe to make sure that part of the manufacturing
>>> process is done correctly and that an Ethernet address is intentionally
>>> assigned.
>>
>>
>> There is no feature like that as far as I know. Can you not just read
>> back the ethaddr variable and compare it to what you intended to
>> program?
>
>
> I don't understand what you are suggesting, sorry.
>
> A long time ago I wrote code
>
> /*
>  * This is an invalid MAC address.
>  * Having it be invalid prevents connection to
>  * the network until the U-Boot "setenv ethaddr"
>  * command is run, which selects a valid MAC
>  * Refer to the section titled "Null values" in
>  * http://standards.ieee.org/regauth/oui/tutorials/UseOfEUI.html.
>  */
> #define CONFIG_ETHADDR      ff:ff:ff:ff:ff:ff
>
> The referenced section says, "... The all-zeros and all-ones EUI-64 values,
> 00-00-00-00-00-00-00-00hex and FF-FF-FF-FF-FF-FF-FF-FFhex, respectively, are
> owned by the IEEE Registration Authority and will never be assigned, and are
> invalid for use as identifiers. ..." (which doesn't really apply to my case,
> but I was hopeful that I could generalize to my case)
>
> At the time I wrote this code, I was trying to cope with a mfg process that
> was about to ship systems without setting correct MAC addresses. I was new
> to U-Boot at that time but found that U-Boot would not build without
> CONFIG_ETHADDR defined.
>
> I subsequently checked again and found that U-Boot connects to the network
> regardless of the value of CONFIG_ETHADDR and even when it is not defined at
> compile-time or run-time.
>
> I left it there, instead I made changes to the mfg process so that "setenv
> ethaddr" was more likely to always get done.

I was simply suggesting that you add a step to your MFG that reads
back the ethaddr after rebooting the target to ensure that the MAC was
programmed.

> I would have liked a way to configure the default U-Boot environment so that
> no network connection could be made until a MAC address had been set, but I
> wasn't willing to spend any more time figuring out how to make that happen
> (possibly by submitting a patch upstream)

You can get that behavior simply by not defining CONFIG_NET_RANDOM_ETHADDR.

#ifdef CONFIG_NET_RANDOM_ETHADDR
                net_random_ethaddr(pdata->enetaddr);
                printf("\nWarning: %s (eth%d) using random MAC address - %pM\n",
                       dev->name, dev->seq, pdata->enetaddr);
#else
                printf("\nError: %s address not set.\n",
                       dev->name);
                return -EINVAL;
#endif

Cheers,
-Joe