[openib-general] OpenSM doesn't start on p5 570

[Hal Rosenstock]

Hi,

On Mon, 2006-01-16 at 08:44, Hal Rosenstock wrote:
> On Mon, 2006-01-16 at 06:56, Andrey Slepuhin wrote:
> > On Mon, Jan 16, 2006 at 06:31:46AM -0500, Hal Rosenstock wrote:
> > > Hi,
> > > 
> > > On Mon, 2006-01-16 at 05:48, Andrey Slepuhin wrote:
> > > > Dear folks,
> > > > 
> > > > I have a problem starting opensm on a p5 570 machine.
> > > 
> > > Is this the first time trying this on a p5 machine ?
> > > 
> > 
> > Yes.
> > 
> > 
> > > > Jan 16 13:30:55 737939 [43027B20] -> umad_receiver: ERR 5413: Failed to obtain request madw for received MAD(method=0x81
> > > > attr=0x11) -- dropping
> > > 
> > > This means that no matching transaction was found in transaction match
> > > table. This may be an endian problem with the tid.
> > > 
> > > Can you validate the tid (print them out) in both get_madw and put_madw
> > > in osm_vendor_ibumad.c ? Since this seems to happen early on, there
> > > shouldn't be too many of these. Thanks.
> > 
> > I got the following:
> > 
> > put_madw: tid=0x1234
> > get_madw: tid=0x1b00001234
> 
> This looks like an endian issue. I will have a patch for you to try
> later. Stay tuned. Thanks.

Can you try this patch and let me know if this works for you ? Thanks.

-- Hal

Index: osm_vendor_ibumad.c
===================================================================
--- osm_vendor_ibumad.c (revision 5016)
+++ osm_vendor_ibumad.c (working copy)
@@ -137,7 +137,7 @@ static osm_madw_t *
 get_madw(osm_vendor_t *p_vend, ib_net64_t *tid)
 {
        umad_match_t *m, *e;
-       ib_net64_t mtid = (*tid & 0xffffffff00000000llu);
+       ib_net64_t mtid = (*tid & cl_ntoh64(0x00000000ffffffffllu));
 
        cl_spinlock_acquire( &p_vend->match_tbl_lock );
        for (m = p_vend->mtbl.tbl, e = m + p_vend->mtbl.max; m < e; m++) {


> > > Are the two HCAs on separate machines ?
> > 
> > No, at the moment they are on the same machine.
> 
> You should be able to run this in loopback. I have done this. Just
> wondering about the topology just to be sure...
> 
> -- Hal
> 
> > Best regards,
> > Andrey
> 
> _______________________________________________
> openib-general mailing list
> openib-general at openib.org
> http://openib.org/mailman/listinfo/openib-general
> 
> To unsubscribe, please visit http://openib.org/mailman/listinfo/openib-general