On Mon, Mar 09, 2009 at 12:25:12PM +0100, Arnoud Vermeer wrote: > We commented out the following lines, to test if it is indeed an > End-of-RIB-marker that is acting up, and it turns out it isn't. > > in rde.c line 2613 we commented out this: > > if (peer->capa_received.restart && peer->capa_announced.restart) > peer_send_eor(peer, afi, safi); > > This is the only place where the peer_send_eor function is called, and > commented out, the bug remains. Hence we assume it is not an eor message > that causes the issue... but an update generated somewhere else. > > Because the empty update is sent out to all connected parties, I think > it has something to do with the 'announce all' capability. >
yes, I had a quick mail exchange with henning about that. There seems to be a wild update that causes this bad updates. I'm currently in Japan preparing everything for AsiaBSDCon plus some traveling. As soon as I can get my head free of all the rest I will look into it. I have a few ideas but nothing was obvious enough to be seen by glancing over the code. Btw. does this only happen with full IPv6 feeds or are a few announcements already enough? -- :wq Claudio
