On Wed, 2005-03-02 at 20:40 -0800, Roland Dreier wrote: > Tom> # arping -c 2 -w 3 -D -I ib0 192.168.0.62 > Tom> ARPING 192.168.0.62 from 0.0.0.0 ib0 > Tom> Sent 2 probes (2 broadcast(s)) > Tom> Received -1 response(s) > > What are the network startup scripts expecting? How is arping getting > so confused that it reports -1 responses? (Sorry, haven't had a > chance to look at the code).
That is a good question. I think the code is b0rked. -1 is coming from the "received" variable. This is never initialized. Initializing it to 0 at least causes arping to fail gracefully, letting ifup continue. I have opened fedora bug 150156 regarding this and emailed the arping maintainer. > Tom> I have looked at the arping code, and it seems to be crafting > Tom> the packet correctly, even using 32 in the type field, so I > Tom> am a bit befuddled as to why this isn't working. > > What packet does arping create? Unfortunately, because a "normal" > IPoIB packet doesn't include any encapsulation beyond the 4 bytes of > ethertype/reserved, it's a little difficult for userspace to send > broadcast packets. If arping tries to create an ethernet-like header, > then the IPoIB driver is going to get a little confused. OK, well I thought it was right according to the ARP IPoIB encapsulation IETF draft. I will look at it a bit more in depth tomorrow. In any event, what is the right format of the packet for userspace to craft? Thanks, -tduffy
signature.asc
Description: This is a digitally signed message part
_______________________________________________ openib-general mailing list [email protected] http://openib.org/mailman/listinfo/openib-general To unsubscribe, please visit http://openib.org/mailman/listinfo/openib-general
