Comment 3 for bug 917817

I am more confident in thinking Maverick and before is ok. However I think this is because there certain markers are reset at the beginning of the loop. The rds_cong_update_alloc should actually initialize a message that is sized to have the congestion map in it. It seems the problem with the new code is that maybe there is a case where the data offset is not reset to zero before the update congestion map message is sent. Then xmit returns the size of the message but the code thinks there was something already done...