Date: 09 Jan 1998 14:03:37 +0100 From: sperber@informatik.uni-tuebingen.de (Michael Sperber [Mr. Preprocessor]) To: dg@root.com Cc: questions@freebsd.org Subject: Re: Networking problem: RPC: No buffer space available Message-ID: <y9liurtg5p2.fsf@modas.informatik.uni-tuebingen.de> In-Reply-To: David Greenman's message of "Fri, 09 Jan 1998 03:07:50 -0800" References: <199801091107.DAA04287@implode.root.com>
next in thread | previous in thread | raw e-mail | index | archive | help
Many thanks for the prompt response! >>>>> "David" == David Greenman <dg@root.com> writes: >> I'm having severe network throughput problems on a 2.2.5 box with an >> NE2000 clone. There's no excessive packet loss or anything else I can >> pinpoint it to. However, there's one reproducible: >> >> lula[35] spray modas >> sending 1162 packets of lnth 86 to modas ...spray: RPC: Unable to send; errno = No buffer space available >> >> Can anyone comment on this? David> "ENOBUFS" (No buffer space available) would be the proper response and is David> expected. Spray just blasts out a pile of (UDP?) packets, without using any David> sort of flow/congestion control. As soon as the interface output queue limit David> is reached (50 packets), the system will start to discard them and return the David> ENOBUFS error. Ah, I see. So there the output queue must be filling up. David> Can you describe the symptoms of your 'excessive packet loss'? As I said, there isn't any. Here's a more detailed description of my setup: I have two FreeBSD hosts in a local Ethernet: Name: lula.informatik.uni-tuebingen.de Address: 134.2.12.20 ... this one's the desktop PC with the NE2000 board running 2.2.5. Name: loosimausi.informatik.uni-tuebingen.de Address: 134.2.13.18 ... a laptop with a D-Link Ethercard running 2.2.2/PAO. The netmask is 255.255.248.0, and neither is running routed. Neither netstat, nor traceroute, nor route is reporting anything unusual between the two. However, throughput lula -> loosimausi is consistently abysmal (around 8k/minute) whereas throughput loosimausi -> lula is consistenly good. Throughput between any of these two and any other machine on the network is consistently good. The only visible thing unusual is that, when a data is being transferred from lula -> loosimausi, spray'ing from lula anywhere reports the above problem. loosimausi doesn't ever, as far as I can tell. -- Cheers =8-} Mike Friede, Völkerverständigung und überhaupt blabla
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?y9liurtg5p2.fsf>