Date: Wed, 2 Oct 1996 19:58:06 +1000 (EST) From: Rowan Crowe <rowan@ozramp.net.au> To: David Greenman <dg@root.com> Cc: bugs@FreeBSD.org Subject: Re: Possible low level TCP bug Message-ID: <Pine.LNX.3.91.961002194920.3654A-100000@ozramp> In-Reply-To: <199610020907.CAA06730@root.com>
next in thread | previous in thread | raw e-mail | index | archive | help
On Wed, 2 Oct 1996, David Greenman wrote: > >> >The behaviour is this: seemingly small amounts of complete data (ie: a > >> >line on IRC) are split into two packets, and are sent about a second > >> >apart. The trace from my current router (not a FreeBSD box) shows this > >> >behaviour. Note the first packet has the PSH attribute. > >> > >> What kind of ethernet card are you using? ...or is this with PPP or SLIP? > > > >Don't laugh: I use 3c501 ethernet cards. The capture I sent to you was > >carried over the ethernet from my FreeBSD box. However, I doubt that's > >the problem, as I've observed inbound data (via serial/PPP) with this > >same quirk. > Actually, it's quite likely that the 3c501 could be the problem. The card > has a total of 2K bytes of buffering - not enough to both receive and transmit > a packet. Of all of the cards supported in FreeBSD, I'd rate the 3c501 at > the very bottom in all respects. This isn't the first complaint like this I've > seen with the 3c501. You'd do well to find an 8bit 3c503 or WD8003. As for > PPP, that could be caused by a different problem. More analysis is needed. I'm well aware of the downfalls of using a 3c501, however they were given to me for free, and most of the use is to link my other 2 machines on the LAN with the outside world via a 14.4k link on the router. There's not a lot of high traffic at all. I've had surprisingly good results with them. If I had more than 3 cards on the network I'm sure things wouldn't work quite so well. :-) (BTW, all my machines are AT or better, I just happened to have the 3c501s). Anyway, as shown by my 2 captures, *exactly* the same problem is occuring, regardless of whether the packet is transported by the 3c501. Note that both times the split occured at the same point, that is, once when generated by my FreeBSD box (and carried over the 3c501s), and the other generated by a remote FreeBSD box coming through a PPP link, no 3c501 or local FreeBSD box involved. Thanks for the info on tracing. :) Cheers.
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?Pine.LNX.3.91.961002194920.3654A-100000>