[openib-general] Solaris IPoIB MTU with OpenSM

[Nitin Hande]

Hal,


On Thu, 2005-02-17 at 13:12, Hal Rosenstock wrote:
> Hi Nitin,
> 
> On Wed, 2005-02-16 at 17:33, Nitin Hande wrote: 
> > On Wed, 2005-02-16 at 13:26, Hal Rosenstock wrote:
> > > On Wed, 2005-02-16 at 16:08, Nitin Hande wrote:
> > > > Hal,
> > [snip..]
[snip...]
> > > > 
> > > > 
> > 
> > Before the patch the selector was coming 04. Do you reply 84 seeing a
> > particular component mask and otherwise 01 ??(I think not..) 
> 
> I agree that OpenSM responds/should respond the same regardless of the
> component mask in the request.
> 
> I was unaware of OpenSM responding with MTU of 01 until now. I have a
> theory as to this. Any chance I can get the osm logs from a -V run of
> the above ? 
> 
> I also have a simple patch below to try which is just to test the
> theory. This is off the latest version but should be easy to apply to
> any version of osm_sa_mcmember_record.c.
> 
> This is separate from the support for PathRecords with multicast DGID
> and/or DLID. I have the changes for this scoped out and should be able
> to implement by early next week.

So I tried the latest patches and preliminarily things seem to be
working fine. The PathRecord response is successful and so is the MTU
correct. I need to spend some more time looking at MAD and confirm it. I
could configure both interfaces and ping each other this time. Here is
some out on the solaris side:

[root at dongon ~]# ifconfig -a
lo0: flags=2001000849<UP,LOOPBACK,RUNNING,MULTICAST,IPv4,VIRTUAL> mtu
8232 index 1
        inet 127.0.0.1 netmask ff000000 
ibd0: flags=1000843<UP,BROADCAST,RUNNING,MULTICAST,IPv4> mtu 2044 index
28
        inet 192.168.100.105 netmask ffffff00 broadcast 192.168.100.255
        ipib 0:2c:0:16:fe:80:0:0:0:0:0:0:0:2:c9:1:9:76:51:d1 
.....
[root at dongon ~]# ping -s 192.168.100.104
PING 192.168.100.104: 56 data bytes
64 bytes from 192.168.100.104: icmp_seq=0. time=0.590 ms
64 bytes from 192.168.100.104: icmp_seq=1. time=0.434 ms
64 bytes from 192.168.100.104: icmp_seq=2. time=0.365 ms

the other side is a openib interface runing OpenSM.

So after this test above, I try to run snoop on the solaris interface
and get the following error message from the layer below IPoIB:

Feb 22 19:50:25 dongon.SFBay.Sun.COM ibd: [ID 517869 kern.info] NOTICE:
ibd0: HCA GUID 0002c901097651d0 port 1 PKEY ffff Could not get list of
IBA multicast groups

My preliminary assumption is that OpenSm is not returning the list of
multicast groups that the ibd interface has joined. I will look at the
MAD's tomorrow and try to ascertain that.

Thanks
Nitin


 



> 
> Thanks.
> 
> -- Hal
> 
> Index: osm_sa_mcmember_record.c
> ===================================================================
> --- osm_sa_mcmember_record.c	(revision 1821)
> +++ osm_sa_mcmember_record.c	(working copy)
> @@ -1325,11 +1325,13 @@
>    /* copy qkey mlid tclass pkey sl_flow_hop mtu rate pkt_life
> sl_flow_hop */
>    __copy_from_create_mc_rec(&mcmember_rec, &p_mgrp->mcmember_rec);
>  
> +#if 0
>    if(p_mgrp->well_known)
>    {
>      p_mgrp->mcmember_rec.mtu = mtu;
>      mcmember_rec.mtu = mtu;
>    }
> +#endif
>  
>    /* Release the lock as we don't need it. */
>    CL_PLOCK_RELEASE( p_rcv->p_lock );
> 
> 
> 
> _______________________________________________
> 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