Design proposal on removing /org/openbmc/settings/boot_policy"

[vishwa]

On 08/17/2017 10:19 PM, Deepak Kodihalli wrote:
> On 17/08/17 9:03 pm, vishwa wrote:
>
>> One of the comment from Deepak in the review was : "If something is
>> ONETIME, then should we save the corresponding Mode / Source to the
>> Flash and I said the answer was "Yes". I see what is being said here in
>> the email is different from what the comment was before.
>>
>> If a setting is ONETIME, then the host comes down and sets the property
>> to DEFAULT and does not do anything with PERMANENT.
>>
>> https://github.com/open-power/petitboot/blob/master/discover/platform-powerpc.c#L874 
>>
>>
>>
>> I don't think so we are introducing any new behavior here. All that was
>> done before is being moved to this new D-Bus object.
>> If something needs to be treated PERMANENT, then it needs to be set by
>> the user as it was before.
>
> Vishwa, I suppose the concern is that we're relying on a specific host 
> behavior, and we're also assuming the host will perform certain 
> actions as per the IPMI specification. So all this is fine as long as 
> these assumptions hold, but they may not always.
>
> In other words, the BMC must not assume it's safe to overwrite a 
> persisted permanent setting with a onetime setting in the assumption 
> that a host will overwrite that value again with a default/permanently 
> desired setting.

I don't think so we are coding anything here that is tightly bound to 
anything. Host is not making anything PERMANENT based on anything. So if 
something needs to be permanent, it needs to be set by the user. It's 
just that on seeing ONETIME, the host sets the flag to DEFAULT and that 
makes perfect sense since this setting is for the Host and post 
consuming it, the consumer would need to say that its been consumed and 
hence reset to DEFAULT.

>
> Regards,
> Deepak
>