[openib-general] RE: [PATCH ] osmtest general cleanups

[Liran Sorani]

Hi , Hal.
The spec I'm referring to is IB Spec Release 1.2 (Volume 1) , p.754 in the
13.5.1 section , look at the third paragraph and find the following quote : 
" Note that it is not required by IBA that GS managers use QP1 as the
source QP used to send management packets to GS agents. GS managers
may send packets from any QP other than QP0."

The InformInfo test flow sends a Trap (using QP0) , this part can be
implemented by umad . In response the SM should send a Report ( to a
previously subscribed node).

-----Original Message-----
From: Hal Rosenstock [mailto:halr at voltaire.com]
Sent: Tuesday, August 09, 2005 3:53 AM
To: Liran Sorani
Cc: openib-general at openib.org; Amit Krig
Subject: RE: [PATCH ] osmtest general cleanups


Hi Liran,

On Mon, 2005-08-08 at 09:58, Liran Sorani wrote: 
> Hi ,
> Regarding the inform_info test flow , the reason I need (possibly)
> ib_verbs is to subscribe a notice report through a QP other then QP1
> (permitted by IB Spec see p754 , 13.5.1) ,

What version of the spec are you referring to ? Can you make the
reference more specific ? Is it 13.5.1.2?

>  generate a trap , then validate the received report through that QP.
> To enable Set/Recieve notice mads through QP other then QP1 , I think
> , I need the ib_verbs

While the subscription (InformInfo) could be through a user SA client,
the Trap/Notice could be direct to some QP other than QP1 and hence
would require uverbs for at least this part.

Do these tests cause SM generated events (ports in and out of service,
multicast groups, etc.) ?

-- Hal

> Regarding the rest , will be fixed in a seperate patch .
> 
> 
> --Original Message--
> From: Hal Rosenstock [mailto:halr at voltaire.com]
> Sent: Monday, August 08, 2005 3:32 PM
> To: Liran Sorani
> Cc: openib-general at openib.org; Amit Krig
> Subject: RE: [PATCH ] osmtest general cleanups
> 
> 
> Hi Liran,
> 
> On Sun, 2005-08-07 at 11:28, Liran Sorani wrote:
> > Hi , Hal.
> > I've a few minor fixes (several white space glitches) , pls take the
> > attached file , instead of the previous  ,
> 
> Thanks. Applied with some minor modifications and caveats below.
> 
> In the future can your patches be submitted as text rather than
> attachments ? That is the norm for doing this.
> 
> Some comments below.
> 
> > thanks  Liran.
> > 
> > --Original Message--
> > From: Liran Sorani 
> > Sent: Sunday, August 07, 2005 5:57 PM
> > To: 'halr at voltaire.com'
> > Cc: 'openib-general at openib.org'; Amit Krig
> > Subject: [PATCH ] osmtest general cleanups
> > 
> > 
> > Hi , Hal.
> > The attached patch should be applied to osmtest repository.
> > It contain several cleanups (on most of the files)   :
> 
> It is easier if there is a patch per idea rather than an amalgam.
> 
> > -Removal of inform info flow .
> > -Unique error messages
> 
> There are some non real error message numbers (neither hex nor
> decimal)
> in osmt_multicast.c. Also, osmtest.c and osmt_service.c still have
> duplicates.
> 
> > -Makefile.am update for compilation (required osm_helper object)
> 
> osm_helper is part of the libopensm since r2973 so this part of the
> patch was not applied. Please update your osm directory.
> 
> > -Remove vendor dependencies.
> 
> Remove some OSM_VENDOR_INTF_MTL vendor dependencies
> (There are still some in osmt_slvl_vl_arb.c).
> 
> > The inform info flow should be carefully ported since it requires
> > direct access to ib_umad (possibly ib_verbs too) .
> 
> umad would be a temporary measure. 
> 
> Why would ib_uverbs direct access be needed ?
> 
> >  I'll send another patch for it next week .
> 
> Thanks.
> 
> -- Hal
> 
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openfabrics.org/pipermail/general/attachments/20050809/bf2740b1/attachment.html>