[openib-general] RE: [PATCH 4/4] opensm: no need to wait for pkey_mgr

[Eitan Zahavi]

Hi Sasha,

So how would the partition manager signal to the LID manager that some
transaction were made and needed to be waited on?

Normally a manager reports back the need for waiting in its return code
that is assigned as the next signal to the state machine. 

So if this signal is not provided from the partition manager (which by
your patch returns void) how would the state machine actually wait for
it if the LID manager returns OSM_SIGNAL_DONE ?

Although this can be fixed by forwarding the result of the partition
manager to affect the LID manager next signal calculation I would rather
not do it. Keeping the state machine as "linear" as possible have great
merits in avoiding extra complexity and bugs.

Eitan Zahavi
Design Technology Director
Mellanox Technologies LTD
Tel:+972-4-9097208
Fax:+972-4-9593245
P.O. Box 586 Yokneam 20692 ISRAEL


> -----Original Message-----
> From: Sasha Khapyorsky [mailto:sashak at voltaire.com]
> Sent: Thursday, May 04, 2006 3:52 AM
> To: Eitan Zahavi
> Cc: Hal Rosenstock; openib-general at openib.org; Ofer Gigi; Yael Kalka
> Subject: Re: [PATCH 4/4] opensm: no need to wait for pkey_mgr
> 
> Hello Eitan,
> 
> On 11:55 Tue 02 May     , Eitan Zahavi wrote:
> >
> > This patch breaks the basic concept for OpenSM to report "SUBNET UP"
> > only if all SubnMgt.Set were sent successfully. I do not think we
want
> > to remove this feature.
> 
> This patch does not break the basic concept (or just in case if there
> are bugs). Instead, pkey updater after sending all set requests will
let
> for the next OpenSM resweep component to run (specifically it is LID
> assignment manager). Then final WAIT state will "collect" responses
from
> both components.
> 
> I think my subject was unclear, and better name would be "parallel
pkey
> and lid managers execution" or like this.
> 
> Of course such "parallelism" is possible only with "independent"
> components (may be more than two BTW), where functionality of one
> component does not depend on results of other component's work.
> 
> So actual question with this patch is could "pkey manager" and "lid
> assignment" run in parallel?
> 
> Sasha.