Re: Code levels in error log entries

[Matt Spinler <mspinler@linux.vnet.ibm.com>]

On 2018-04-30 12:31, Deepak Kodihalli wrote:
> On 30/04/18 8:18 pm, Matt Spinler wrote:
>> On 2018-03-21 12:23, Matt Spinler wrote:
>>> For IBM systems, we need to add the current code level to each error 
>>> log
>>> object as it's created to make it easier on the lives of support
>>> personnel.  It would be a new interface (or maybe use an existing
>>> code version one) on the /xyz/openbmc_project/logging/entry/X 
>>> objects.
>>> 
>>> Does anyone here that also uses error logs have an opinion on if they
>>> would also be interested in this or not, so it would go into the 
>>> common
>>> logging code?
>>> 
>>> Unless others are interested in it, I plan on putting it in an IBM
>>> specific repository.  (or maybe turn on with a configure option?)
>> 
>> 
>> After some additional discussion, we decided this would be useful 
>> enough to
>> warrant putting it into the base error logging code.  I plan on 
>> adding the
>> xyz.openbmc_project.Software.Version interface to the logging objects 
>> and
>> populating it with the VERSION_ID field from /etc/os-release.
> 
> I see the Version interface also has a 'VersionPurpose' enum property.
> What would you populate that with? VersionPurpose.BMC? What if the
> error was from host and the version required was that of the host
> firmware?

I was always planning on putting the BMC code version into this.  In the 
case of
OpenPower, the host already includes their code version in their 
embedded PEL error
I believe.

> 
> Also, the value that gets into the /etc/os-release is known to the
> os-release recipe, so you could provide that to the logging recipe at
> build time (ipmi does something similar).


Thanks.  Will look into that.


> 
>> 
>> 
>> 
>>> 
>>> Thanks
> 
> Thanks,
> Deepak