[ofa-general] Re: OpenSM IPv6 consolidation

[Hal Rosenstock]

On Thu, 2008-05-29 at 14:35 -0700, Ira Weiny wrote:
> On Thu, 29 May 2008 12:27:12 -0700
> Hal Rosenstock <hrosenstock at xsigo.com> wrote:
> 
> > Ira,
> > 
> > In osm_sa_mcmember_record.c:__search_mgrp_by_mgid, there is:
> > 
> > #define PREFIX_MASK (0xff12601b00000000ULL)
> > 
> > Shouldn't all scopes be consolidated so this should be:
> > 
> > #define PREFIX_MASK (0xff10601b00000000ULL)
> > 
> > or was this intentional for some reason ?
> > 
> 
> It seemed reasonable for this to consolidate link-local only because according
> to my IPv6 book, solicited node multicast is the particular range,
> ff02::1:ff00:0/104
> 
> However, I am a bit confused about how the scope bits map from the IP address
> to the MGID.  The MGID refers only to the IB-subnet scope _not_ IP, therefore
> what I said above might not matter because we are now talking about the IB
> scope.

Right; the IP scope is different from the IB scope.

> But that begs the question: Can a node issue an SNM request to a node in
> another IB subnet? 

Yes, as IPoIB subnets can span IB subnets.

-- Hal

>  (I think the answer is yes if the IP subnet spans more than
> one IB subnet)  In that case, the SNM address would be in the range
> ff02::1:ff00:0/104 but what MGID would that map onto in IB?  I think the
> current mapping results in an IB link-local scope.  So would a router have to
> forward it even though the IB scope is link-local?

> Now my head hurts...  :-(
> 
> Ira
>