[U-Boot] [PATCH 05/15] powerpc/83xx/km: add MV88E6122 switch support for kmvect1

[Holger Brunck]

On 01/18/2013 10:12 PM, Kim Phillips wrote:
> On Fri, 18 Jan 2013 10:14:39 +0100
> Holger Brunck <holger.brunck at keymile.com> wrote:
> 
>> On 01/18/2013 01:44 AM, Kim Phillips wrote:
>>> On Thu, 17 Jan 2013 13:47:51 +0100
>>> Holger Brunck <holger.brunck at keymile.com> wrote:
>>>
>>>> +	if (piggy_present()) {
>>>> +		uchar enetaddr[6];
>>>> +		if (eth_getenv_enetaddr("ethaddr", enetaddr)) {
>>>> +			/* increment last 3 bytes */
>>>> +			u32 val = (enetaddr[3] << 16) + (enetaddr[4] << 8)
>>>> +				+ enetaddr[5];
>>>> +			val++;
>>>> +			enetaddr[3] = (val >> 16) & 0xFF;
>>>> +			enetaddr[4] = (val >>  8) & 0xFF;
>>>> +			enetaddr[5] = (val >>  0) & 0xFF;
>>>> +			if (!eth_setenv_enetaddr("eth1addr", enetaddr)) {
>>>> +				setenv("ethact", "UEC2");
>>>> +				setenv("netdev", "eth1");
>>>> +				puts("using PIGGY for network boot\n");
>>>> +			} else
>>>> +				puts("using frontport for network boot\n");
>>>> +		}
>>>> +	} else
>>>> +		puts("using frontport for network boot\n");
>>>> +#endif
>>>> +
>>>
>>> AFAIK, assigning the switch MAC address to the front port's MAC
>>> address plus one isn't acceptable practice for u-boot board code.
>>> Get eth1addr in the same manner ethaddr is obtained/assigned.
>>
>> Because you may get a MAC adress which is already given to someone else? In our
>> case this can't happen. We reserve for one instance of this board exactly two
>> MAC adresses. The first one is read out at startup from the inventory in
>> keymile/common/ivm.c. The second one is calculated like above. But it is a
>> unique MAC adress, because exactly this MAC adress is reserved for this board.
> 
> it can happen if if setting ethaddr from eeprom somehow fails, and/or
> the user manually modifies ethaddr in the environment.
> 
>> But maybe we should move the code to the location where we set ethaddr,
> 
> yes, indeed
> 

ok I'll move it to this location. Then it's definetely based on the value read
from ivm. And it's then easy to check for errors if e.g. the IVM is not readable.

Regards
Holger