public inbox for linux-kernel@vger.kernel.org
 help / color / mirror / Atom feed
* tcp connections hanging at connect
@ 2001-08-02 14:18 Philipp Reisner
  0 siblings, 0 replies; only message in thread
From: Philipp Reisner @ 2001-08-02 14:18 UTC (permalink / raw)
  To: linux-kernel

Hi,

I have discovered here something that causes TCP connections to 
hang if one of the initial packets is lost.

We have a script which runs scp every 10 seconds, and the script
simply hangs after some minutes to an hour. The script runs on 
a Linux-2.2.19 Box (we have also tested Linux-2.4.2) and the ssh server
is running on some Windows box.

Here is the good case:

15:26:11.100413 192.168.53.4.4819 > 212.31.78.62.22: S 2626815412:2626815412(0) win 16060 <mss 1460,sackOK,timestamp 809461847[|tcp]> (DF)
15:26:11.119964 212.31.78.62.22 > 192.168.53.4.4819: S 3560917622:3560917622(0) ack 2626815413 win 17520 <mss 1460,nop,wscale 0,nop,nop,timestamp[|tcp]> (DF)
15:26:11.120011 192.168.53.4.4819 > 212.31.78.62.22: . ack 1 win 16060 <nop,nop,timestamp 809461848 0> (DF)
15:26:11.228046 212.31.78.62.22 > 192.168.53.4.4819: P 1:24(23) ack 1 win 17520 <nop,nop,timestamp 6062108 809461848> (DF)

Here is the hang:

12:01:24.753703 192.168.53.4.4442 > 212.31.78.62.22: S 2538486974:2538486974(0) win 16060 <mss 1460,sackOK,timestamp 808233194[|tcp]> (DF)
12:01:24.798610 212.31.78.62.22 > 192.168.53.4.4442: S 3871618076:3871618076(0) ack 2538486975 win 17520 <mss 1460,nop,wscale 0,nop,nop,timestamp[|tcp]> (DF)
12:01:24.798729 192.168.53.4.4442 > 212.31.78.62.22: . ack 1 win 16060 <nop,nop,timestamp 808233198 0> (DF)
12:01:28.048197 212.31.78.62.22 > 192.168.53.4.4442: S 3871618076:3871618076(0) ack 2538486975 win 17520 <mss 1460,nop,wscale 0,nop,nop,timestamp[|tcp]> (DF)
12:01:34.611132 212.31.78.62.22 > 192.168.53.4.4442: S 3871618076:3871618076(0) ack 2538486975 win 17520 <mss 1460,nop,wscale 0,nop,nop,timestamp[|tcp]> (DF)

192.168.53.4: Is the Linux box.
212.31.78.62: Is the Windows box.

It looks like that the packet at 12:01:24.798729 never reaches the Windows
box. Ok -- That is probabely why the Windows box resends it's syn packet
(at 12:01:28.048197 and 12:01:34.611132). 

BTW, the Linux box is convinced that the connection is established 
(confirmed with lsof), while the Windows box probabely does not think
that there is a connection.

The question is, why is Linux not responding to the resent syn packets ?

PS: the process (scp/ssh) on the Linux side of the connection wants
    to read from the socket (confirmed with strace).

-Philipp

^ permalink raw reply	[flat|nested] only message in thread

only message in thread, other threads:[~2001-08-02 14:18 UTC | newest]

Thread overview: (only message) (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2001-08-02 14:18 tcp connections hanging at connect Philipp Reisner

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox