[ofa-general] SubnAdmGet (6777)

[Hal Rosenstock]

On 6/1/09, Sean Hefty <sean.hefty at intel.com> wrote:
>>Yes, it is different from GetTable in that SA pares the responses down
>>to that but Get doesn't (have that additional language to pare them
>>down).
>
> This seems like an implementation issue (aka bug) with the SA to me.

No; it's what the spec says.

>>The language about NumbPath for Get was originally added to indicate
>>that the NumbPath was ignored on a Get even if it was included in the
>>component mask.
>
> It states that it's ignored and a value of 1 is used.  What else would a
> NumbPath value of 1 mean if it's completely ignored?  I consider this a spec
> bug.  :)

It was also to deal with the broken case of Get PR with NumbPath > 1.

> From an implementation view, requiring users to use SubnAdmGetTable to get a
> single path record is less efficient than returning a single PR from
> SubnAdmGet.

Yes, RMPP is an overhead when the response is a single MAD but is this
significant ? Anyhow, how can the spec be changed in a way that
doesn't break existing implementations ?

-- Hal

> How have other SM implementations (not based on opensm) interpreted NumbPath
> for PR SubnAdmGet?
>
>