[openib-general] [PATCH] osm: fix num of blocks of GUIDInfo GetTable query

[Eitan Zahavi]

Hi Hal,

When is a complete fix expected?
Meanwhile osmtest on large enough cluster is not passing due to the huge
number of GUID blocks...

If this full fix not anticipated soon can we have the simple fix applied
first?

Eitan Zahavi
Senior Engineering Director, Software Architect
Mellanox Technologies LTD
Tel:+972-4-9097208
Fax:+972-4-9593245
P.O. Box 586 Yokneam 20692 ISRAEL


> -----Original Message-----
> From: Hal Rosenstock [mailto:halr at voltaire.com]
> Sent: Saturday, June 10, 2006 9:11 PM
> To: Eitan Zahavi
> Cc: OPENIB
> Subject: Re: [PATCH] osm: fix num of blocks of GUIDInfo GetTable query
> 
> Hi Eitan,
> 
> On Sat, 2006-06-10 at 13:12, Eitan Zahavi wrote:
> > Hal Rosenstock wrote:
> > > Hi Eitan,
> > >
> > > On Thu, 2006-06-08 at 07:24, Eitan Zahavi wrote:
> > >
> > >>Hi Hal
> > >>
> > >>I'm working on passing osmtest check. Found a bug in the new
> > >>GUIDInfoRecord query: If you had a physical port with zero
guid_cap
> > >>the code would loop on blocks 0..255 instead of trying the next
port.
> > >
> > >
> > > OK; that's definitely a problem.
> > >
> > >
> > >>I am still looking for why we might have a guid_cap == 0 on some
> > >>ports.
> > >
> > >
> > > PortInfo:GuidCap is not used for switch external ports.
> > >
> > >
> > >>This patch resolves this new problem. osmtest passes on some
arbitrary
> > >>networks.
> > >>
> > >>Eitan
> > >>
> > >>Signed-off-by:  Eitan Zahavi <eitan at mellanox.co.il>
> > >>
> > >>Index: opensm/osm_sa_guidinfo_record.c
> >
>>===================================================================
> > >>--- opensm/osm_sa_guidinfo_record.c	(revision 7703)
> > >>+++ opensm/osm_sa_guidinfo_record.c	(working copy)
> > >>@@ -255,6 +255,10 @@ __osm_sa_gir_create_gir(
> > >>       continue;
> > >>
> > >>     p_pi = osm_physp_get_port_info_ptr( p_physp );
> > >>+
> > >>+    if ( p_pi->guid_cap == 0 )
> > >>+      continue;
> > >>+
> > >
> > >
> > > I think the right fix is to detect switch external ports and use
the
> > > VLCap from port 0 rather than from the switch external port
(unless that
> > > concept is broken in which case it should return 0 records).
> > I think switch external ports do not have any PortGUID assigned to
them since
> > they are not "end port" (i.e. addressable).
> 
> Right; that's what I said earlier in a different way (PortGUID is not
> used for switch external ports).
> 
> > So I think this patch is good enough.
> 
> I think its better (an improvement) but not a complete fix for this
> issue.
> 
> > What if a port reports guid_cap == 0?
> 
> Is that legal ? Shouldn't any port where GUIDCap is valid have a non
> zero GUIDCap ? On any port where GUIDCap is not used (e.g. invalid),
it
> should be ignored.
> 
> > (I understand it is illegal for addressable port
> > but for the SM it is probably better not to assume all ports are
legal...)
> 
> That's my point on what a complete fix for this would include.
> 
> -- Hal
> 
> > EZ
> > >
> > > -- Hal
> > >
> > >
> > >>     num_blocks = p_pi->guid_cap / 8;
> > >>     if ( p_pi->guid_cap % 8 )
> > >>       num_blocks++;
> > >>
> >