I'm seeing the following issue and was hoping for some additional insight: The local IB interface on a "server" node is in CM (connected) mode with an MTU of 64K. dmesg shows numerous errors like: ib0: packet len 8380 (> 2048) too long to send, dropping
Client machines are attempting to access this node with IPoIB UD. When the server ib0 interface is switched back to UD (datagram) mode, this problem goes away but that's only a workaround. Has anyone seen this problem ? Any ideas on what causes this and/or how to resolve it ? I've seen some comments about this occuring prior to path MTU discovery. Is the issue is that some IP services don't utilize path MTU discovery ? -- Hal _______________________________________________ general mailing list general@lists.openfabrics.org http://lists.openfabrics.org/cgi-bin/mailman/listinfo/general To unsubscribe, please visit http://openib.org/mailman/listinfo/openib-general
