Re: Some more 1.3.66 TCP confusion

Thomas =?ISO-8859-1?Q?K=F6nig?= (Thomas.Koenig@ciw.uni-karlsruhe.de)
Wed, 21 Feb 1996 00:16:12 +0100 (MET)


Same bug still applies in 1.3.67, with Linus' little patch from c.o.l.d
applied.

I wrote:

>I just did two "$ socket -w -q linuxbox echo < .newsrc"
>from another, external HP-UX machine, then interrupted them (-w means
>'write only'). This had some bad effects on what 1.3.66 thinks about the state
>of its sockets:

>Active Internet connections
>Proto Recv-Q Send-Q Local Address Foreign Address (State) User
>tcp 49388 51636 myhost.ciw.uni-ka:echo myhp.rz.uni-karls:3220 ESTABLISHED root
>tcp 50644 52732 myhost.ciw.uni-ka:echo myhp.rz.uni-karls:3228 ESTABLISHED root

>On the other hand, the HP box thinks the two connections are in
>FIN_WAIT_1, with 8192 bytes still to be sent. Looking at tcpdump,
>the Linux box sends 'ack 0, win 0' packets on these ports to
>the HP every now and then. The HP box doesn't answer.

-- 
Thomas Koenig, Thomas.Koenig@ciw.uni-karlsruhe.de, ig25@dkauni2.bitnet.
The joy of engineering is to find a straight line on a double
logarithmic diagram.