On Tue, Aug 4, 2009 at 10:03 AM, Hal Rosenstock <[email protected]>wrote:

>
>
>  On Sun, Aug 2, 2009 at 5:45 PM, Todd Rimmer <[email protected]>wrote:
>
>> What is the proper way to control the SLID used for outgoing umad sends?
>>
>> For example, when using LMC>0, the PathRecord returned from the SM for
>> talking to a given remove node may have a SLID which is not the BaseLid for
>> the sender.  How does the sender ensure the correct SLID is used for the
>> outgoing mad?
>>
>> In reviewing the API it seems like the only way to do this is:
>> void *umad = umad_alloc(...);
>>
>> // call various umad calls to initialize address and contents
>> umad_get_mad_addr(umad)->path_bits = lower LMC bits of SLID;
>>
>> umad_send(..., umad, ...);
>>
>> Was path_bits an intentional omission in the API?
>
>
> No; it was an unintentional omission AFAIT.
>
>
>>  It would seem that a function which could update the ib_mad_addr in a
>> umad given a path record would seem appropriate.
>
>
> Seems reasonable to me.
>

On second thought, umad is lower level than this and knows nothing of path
records (that at higher level). Some API like umad_set_addr handling path
bits would be another alternative for this.

-- Hal



>  Care to supply a patch ?
>
> -- Hal
>
>
>>
>> Todd Rimmer
>> Chief Architect
>> QLogic Network Systems Group
>> Voice: 610-233-4852     Fax: 610-233-4777
>> [email protected]  www.QLogic.com <http://www.qlogic.com/>
>>
>> _______________________________________________
>> general mailing list
>> [email protected]
>> http://lists.openfabrics.org/cgi-bin/mailman/listinfo/general
>>
>> To unsubscribe, please visit
>> http://openib.org/mailman/listinfo/openib-general
>>
>
>
_______________________________________________
general mailing list
[email protected]
http://lists.openfabrics.org/cgi-bin/mailman/listinfo/general

To unsubscribe, please visit http://openib.org/mailman/listinfo/openib-general

Reply via email to