[ofa-general] [PATCH] OpenSM/osm_port_info_rcv.c: In __osm_pi_rcv_process_endport, isSMdisabled also indicates that an SM is present so poll SMInfo

[Hal Rosenstock]

On Thu, 2007-04-05 at 13:20, Roland Dreier wrote:
>  > No; the requirements are that it not send SubnSet/Get but still needs to
>  > respond to them.
> 
>  > IsSMdisabled goes in concert with the behavior above. See IBA 1.2 14.4.8
>  > (p. 879).
> 
> OK, I just looked.  Doesn't C14-70 say that SubnGet(SMInfo) _shall_ be
> discarded if IsSMDisabled is asserted (in addition to not sending any
> such queries)?

Yes, I didn't read carefully before I typed...

> I think I'm missing the point of IsSMDisabled.  What I was looking for
> was the motivation behind the spec.  So can you give a quick example
> of some simple situation where setting IsSMDisabled helps me make my
> network work better?

C14-69 describes one such scenario where the SM is disabled out of band
and the only way other SMs know there is an inactive SM there is via
this capmask bit as it will not respond to SMInfo.

-- Hal

>  - R.