[ofa-general] Re: [ewg] APM Example

Hal Rosenstock halr at voltaire.com
Sat Apr 28 07:15:34 PDT 2007


On Sat, 2007-04-28 at 10:11, Abhinav Vishnu wrote:
> Hi Roland,
> 
> On 4/26/07, Roland Dreier <rdreier at cisco.com> wrote:
>             Abhinav> However, VAPI has an event which specifies the
>         successful
>             Abhinav> transition of MIGRATED -> ARMED (I know very
>         well, that
>             Abhinav> it is done through modify_qp). But just the
>         success of 
>             Abhinav> modify_qp does not explicitly tell the time at
>         which the
>             Abhinav> transition successfully occured, does it?
>         
>         You don't know the time that the transition occurred, except
>         that it
>         is between when you called modify QP and when it
>         returned.  But an
>         asynchronous event doesn't really help, does it?  
> 
> It does help. APM is not only defined for network fault tolerance, it
> can
> also be used for load-balancing. With this event, one can know when
> the path is loaded and it is safe to call modify_qp.
> 
> Also, do you have any script which can potentially bring a port down 
> from Active state, without actually unplugging the cable? Please let 
> me know.

ibportstate

-- Hal

> Thanks,
> 
> :- Abhinav
>  
>         All an event would
>         tell you is that the transition occurred some time before the
>         event was 
>         generated, which is some time before when the event was
>         delivered to you.
>         
>             Abhinav> Specifically:
>         
>             Abhinav> VAPI_PATH_MIG_ARMED would make my day. I believe
>         that
>             Abhinav> VAPI_QP_PATH_MIGRATED is similar to 
>             Abhinav> IB_EVENT_PATH_MIG. Please correct me if i am
>         wrong.
>         
>         I see... VAPI_PATH_MIG_ARMED is a new event that was added
>         only in
>         VAPI 4.1.0, which was why I didn't know about it.  Only
>         Mellanox HCAs 
>         support it, it is not specified by the InfiniBand
>         architecture, and I
>         don't really see the point of it (as I tried to explain
>         above).
>         
>         - R.
> 
> 
> 
> -- 
> Abhinav Vishnu 
> Graduate Student
> Computer Science and Engineering
> The Ohio State University 
> 
> ______________________________________________________________________
> 
> _______________________________________________
> general mailing list
> general at lists.openfabrics.org
> http://lists.openfabrics.org/cgi-bin/mailman/listinfo/general
> 
> To unsubscribe, please visit http://openib.org/mailman/listinfo/openib-general




More information about the ewg mailing list