[ofa-general] Re: [PATCH] ibsim: Drop vendor MADs and vendor specific SM MADs

[sebastien dugue]

  Hi Sasha,

On Mon, 27 Jul 2009 18:20:36 +0300
Sasha Khapyorsky <sashak at voltaire.com> wrote:

> On 13:38 Mon 27 Jul     , sebastien dugue wrote:
> > 
> >   This is useful for testing applications that send those kind of MADS.
> 
> How this is useful in such test?

  Because I discovered that when using libibumad, it hangs in umad_recv() for
those MADs when the client is OpenSM regardless of the timeout I set.


> I see that this just drops every sort
> of such MADs without letting chance to client to process it. No?

  Right, not so good, forget it. I guess I'll have to find out why OpenSM
silently drops those MADs and why I get stuck in umad_recv().


  Thanks,

  Sebastien.


> 
> Sasha
> 
> 
> > 
> > Signed-off-by: Sebastien Dugue <sebastien.dugue at bull.net>
> > ---
> >  ibsim/sim_mad.c |   19 +++++++++++++++++++
> >  1 files changed, 19 insertions(+), 0 deletions(-)
> > 
> > diff --git a/ibsim/sim_mad.c b/ibsim/sim_mad.c
> > index 61d4866..b79a5e7 100644
> > --- a/ibsim/sim_mad.c
> > +++ b/ibsim/sim_mad.c
> > @@ -1170,6 +1170,25 @@ int process_packet(Client * cl, void *p, int size, Client ** dcl)
> >  	if ((response = decode_sim_MAD(cl, r, &rpc, &path, data)) < 0)
> >  		return -1;
> >  
> > +	/* Drop vendor MADs */
> > +	if (((rpc.mgtclass >= IB_VENDOR_RANGE1_START_CLASS) &&
> > +	     (rpc.mgtclass <= IB_VENDOR_RANGE1_END_CLASS)) ||
> > +	    ((rpc.mgtclass >= IB_VENDOR_RANGE2_START_CLASS) &&
> > +	     (rpc.mgtclass <= IB_VENDOR_RANGE2_END_CLASS))) {
> > +		IBWARN("ignoring vendor MAD: class 0x%x, attr 0x%x",
> > +		       rpc.mgtclass, rpc.attr.id);
> > +		goto _dropped;
> > +	}
> > +
> > +	/* Drop vendor specific SM MADs */
> > +	if (((rpc.mgtclass == IB_SMI_CLASS) ||
> > +	     (rpc.mgtclass == IB_SMI_DIRECT_CLASS)) &&
> > +	    (rpc.attr.id >= 0xff00)) {
> > +		IBWARN("ignoring vendor specific SM MAD: attr 0x%x",
> > +		       rpc.attr.id);
> > +		goto _dropped;
> > +	}
> > +
> >  	if (rpc.method == 0x7) {
> >  		IBWARN("lid %u got trap repress - dropping", ntohs(r->dlid));
> >  		*dcl = 0;
> > -- 
> > 1.6.3.1
> > 
>