[ofa-general] opensm --lcm

Hal Rosenstock hrosenstock at xsigo.com
Mon Nov 12 12:14:44 PST 2007


On Mon, 2007-11-12 at 22:21 +0200, Sasha Khapyorsky wrote:
> Hi,
> 
> On 19:13 Mon 12 Nov     , Bernd Schubert wrote:
> > 
> > so far we always did run opensm without the --lmc option, but just recently I 
> > read in opensm's manpage about it.
> > 
> > <opensm manpage>
> > This option specifies the subnet's LMC value.  The number of LIDs assigned to 
> > each port  is  2^LMC.   
> > The LMC  value  must be in the range 0-7.  LMC values > 0 allow multiple paths 
> > between ports.  LMC values > 0 should only be used if the subnet topology 
> > actually provides multiple paths between ports, i.e.  multiple
> > interconnects  between  switches.   Without -l, OpenSM defaults to LMC = 0, 
> > which allows one path between any two ports.
> > </opensm manpage>
> > 
> > In one of our configurations we do have a mts2400 (master) switch and 
> > connected to it are two mts14400 switches (each with 3 connections to the the 
> > mts2400).
> > I think from the description of the opensm manpage we need to give the "--lmc" 
> > to make use of the 3 connections from each mts14400 switch to the mts2400 
> > switch, don't we?
> 
> In theory (AFAIR by default OpenSM will not assign LMC for Switch port 0

Non 0 LMC is only valid (per spec) for enhanced SP0. LMC must be 0 for
base SP0.

> due to some chip bug), when target is switch itself. But if you have
> end nodes connected to this switch and targets are end nodes when OpenSM
> will try to use all links for it.
> 
> > Furthermore, from manpage I would think we even need this 
> > option for proper inter-communication between the switch modules?
> 
> For switch modules yes, but again OpenSM will not assign LMC for
> switches (even then --lmc specified), only for end nodes. In order to
> enable it you will need to turn on 'lmc_esp0' option in OpenSM options
> file. But as far as I remember it is not supported by switches now.

It didn't work last time I tried it and I don't think it's been fixed.

-- Hal

> Sasha
> _______________________________________________
> 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 general mailing list