FRU Device type one byte or Two byte Validation

[Kumar Thangavel]

Hi Ed,

    Please find my response below.
    Please search this mail thread "EEPROM Validation issue in Fru Device "
for more information.

Thanks,
Kumar.


On Thu, Nov 25, 2021 at 2:10 AM Ed Tanous <edtanous at google.com> wrote:

> On Wed, Nov 24, 2021 at 1:17 AM Kumar Thangavel
> <kumarthangavel.hcl at gmail.com> wrote:
> >
> > Hi All,
> >
> >        We would like to validate our NIC FRU device type as 8 bit (one
> byte) or 16 bit(two byte).
> >
> >        This device type validation for our FRU is platform specific
> logic.
>
> Please include more details about what this logic is.  In the
> entity-manager design, anything that's truly platform specific needs
> to go into the entity-manager configuration, as all other components
> are common.
>
> The existing logic to get the device type is 8bit or 16bit is not working
exactly for our platform.
Also, writing some bytes and getting device type in eeprom is not safe and
not suggested.
So, we implemented the platform specific logic to get the eeprom device
type in the machine layer.

https://github.com/openbmc/entity-manager/blob/master/src/FruDevice.cpp#L430

> >  This was implemented in the machine layer.
>
> Can you point to it?
>
> The logic is, get the NIC card vendor info from ncsi-netlink command and
based on NIC vendor,
set the eeprom device type as 8bit or 16 bit.

We wanted to interlink this logic in entity-manager.

>
> >        We wanted to interlink this logic with entity-manager code to get
> the device type.
> >
> >        Could you please provide the suggestions on get the device type
> in entity-manager.
>
> Entity-manager provides the exposes records on dbus, for which
> reactors can pick them up and use them.
>

Is there any way to get the dynamic values (device type - 8bit/16bit ) from
the machine layer
to the entity-manager ?

>
> >
> >        Please refer below link for existing devicebit validation in
> entity-manager.
> >
> https://github.com/openbmc/entity-manager/blob/master/src/FruDevice.cpp#L430
>
> I'm not quite following what this has to do with your above, given
> that this isn't intended to be platform specific.
>
>
This existing logic to get the device type 8/16 bit may not work for all
the devices.
We need to find common logic for all the devices without any breakage. But
it is very tricky.

Alternatively, platform specific logic can be an option and interlink with
EM at run time.
We are trying this and requesting suggestions from MAINTAINERS.

>
> > Thanks,
> > Kumar.
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.ozlabs.org/pipermail/openbmc/attachments/20211125/ee727273/attachment-0001.htm>