[ofa-general] Bugs in opensm/libvendor

[Hal Rosenstock]

Sasha,

On Tue, Dec 16, 2008 at 2:43 AM, Sasha Khapyorsky <sashak at voltaire.com> wrote:
> Hi again, Hal,
>
> On 11:03 Mon 15 Dec     , Hal Rosenstock wrote:
>> On Mon, Dec 15, 2008 at 10:38 AM, Sasha Khapyorsky <sashak at voltaire.com> wrote:
>> > On 09:29 Mon 15 Dec     , Mike Heinz wrote:
>> >>
>> >> That's a good question - and I'm going to ask around and double check.
>> >> My first reaction was that you have to specify how many paths you want
>> >> from the query - but you're right, the spec doesn't say that.
>> >
>> > Yes, it looks like this (but I cannot understand "why" :( ).
>>
>> The spec says this (for GetTable) and Gets are requests for 1 path.
>> The reason is to limit the amount of returned path records (and the
>> field limits to 255 records in the response).
>
> Do you know what is a reason for this "127 records" limitation?

Once you get past the scalability discussion (including limiting it to
SGID), is there a need for more than 127 ? I think that allowing more
paths is more important with various other types of wildcarded PR
queries that are "beyond the spec".

-- Hal

>> >But even more
>> > strange (IMHO) limitation is mandatory SGID - actually it should make
>> > illegal such GetTable queries as all-to-all, SLID-to-all, etc.. I
>> > thought that it is permitted.
>>
>> It was decided to force SGID. Neither All to all nor SLID to all by
>> itself are spec'd (you could could add SGID along with SLID to all
>> though). Support for those is a proprietary OpenSM extension which is
>> used for testing at least (and also by saquery command).
>
> Ok. Not a bad extension IMHO :)
>
> Sasha
>