Intel S2600WF (Wolf Pass) BMC MAC addresses issue.

Milton Miller II miltonm at us.ibm.com
Sat Jan 18 09:44:03 AEDT 2020


About 03:29PM 01/17/2020 in some timezone, Johnathan Mantey wrote:
>
>This is expected behavior.
>I believe the MAC address is generated randomly the first time.
>I'm not certain if this will remain the status quo.
>The MAC addresses are pulled from the
>/etc/systemd/network/00-bmc-ethx.network files.
>It is necessary to write the information, and then reboot the BMC for
>the value to take effect.

It's the behavior if the platform is missing some support.

>
>On 1/17/20 1:16 PM, Bruce Mitchell wrote:
>> On my Intel S2600WF (Wolf Pass) the original Purley BMC FW gets its
>MAC
>> addresses properly and consistently,
>> 
>> when reflashing the BMC SPI FLASH the BMC will still get the
>correct and
>> same MAC addresses.
>> 
>>  
>> 
>> Then all I change is using OpenBMC (from here
>>
>https://urldefense.proofpoint.com/v2/url?u=https-3A__github.com_Intel
>-2DBMC_openbmc&d=DwID-g&c=jf_iaSHvJObTbx-siA1ZOg&r=bvv7AJEECoRKBU02rc
>u4F5DWd-EwX8As2xrXeO9ZSo4&m=nAfDq6tUDSiK_KGsv8J_gf4rZRggfe7L3fFnWRBD2
>xQ&s=oiK3NIjr0m1wrqSiymjsJO5bn9SQFCRcKpaOjzAXkss&e= ) and then the
>BMC gets
>> 
>> different MAC addresses each time I reflash the BMC SPU FLASH.
>> 
>> Is this the present expected behavior?
>> 
>> If so will it be corrected to behave like the Purley BMC FW for MAC
>> addresses?
>> 
>>  
>> 
>> Thanks!
>> 
>>  
>> 
>
>-- 
>Johnathan Mantey


Wolf Pass uses the aspeed AST 2500, which does not have space to store
the MAC address inside the SOC.

Das U-boot, the booloader that OpenBMC uses, will write a MAC into the 
hardware if one is configured in the firmware environment ethaddr 
variable.If one is not set, it will generate a random MAC from the 
administratively assigned space.

Likewise, the Linux kernel driver will use a mac programmed into the 
hardware if it is valid, or it too will generate a random MAC from the 
administratively assigned space.

Other OpenBMC systems extract the MAC from a FRU object in the inventory 
and write this to a d-bus variable which then triggers a call to store 
the MAC in the u-boot environment via the fw-setenv utility.

I don't have any specs or hardware for WolfPass so I ccan't comment on 
where the MAC address is stored.  If you have trouble finding the code
to transport the MAC to the U-boot environment I can search for the 
pieces; I know the path exists on romulus and witherspoon platforms in
the meta-openpower and meta-ibm layers.

milton



More information about the openbmc mailing list