TCP connection forwarding troubles

[Phil Karn]

For some time I have routinely websurfed across a forwarded TCP
connection using SSH.  The other end of the TCP tunnel connects to a
Squid proxy cache on the same machine.

This usually works. But I see lots of error messages on each end of
the form

Jun 13 13:22:02 tunnel sshd[32378]: error: channel 0: chan_shutdown_read: shutdown() failed for fd5 [i1 o128]: Transport endpoint is not connected
Jun 13 13:22:11 tunnel sshd[32378]: error: channel 2: chan_shutdown_read: shutdown() failed for fd10 [i1 o128]: Transport endpoint is not connected
Jun 13 13:24:50 tunnel sshd[32378]: error: channel 0: chan_shutdown_read: shutdown() failed for fd5 [i1 o128]: Transport endpoint is not connected
Jun 13 13:25:03 tunnel sshd[32378]: error: channel 1: chan_shutdown_read: shutdown() failed for fd8 [i1 o128]: Transport endpoint is not connected
Jun 13 13:25:03 tunnel sshd[32378]: error: channel 3: chan_shutdown_read: shutdown() failed for fd11 [i1 o128]: Transport endpoint is not connected
Jun 13 13:25:03 tunnel sshd[32378]: error: channel 2: chan_shutdown_read: shutdown() failed for fd10 [i1 o128]: Transport endpoint is not connected
Jun 13 13:25:14 tunnel sshd[32378]: error: channel 5: chan_shutdown_read: shutdown() failed for fd13 [i1 o128]: Transport endpoint is not connected
Jun 13 13:46:02 tunnel sshd[32378]: error: channel 3: chan_shutdown_read: shutdown() failed for fd11 [i1 o128]: Transport endpoint is not connected

This is from the remote end of the tunnel, the machine with the squid
proxy. I see similar messages on the client side.

And occasionally things often just hang. When I kill and restart the
local ssh client that's providing the tunnel, it starts working again.

Both ends are running OpenSSH 2.1.1, Debian Linux 2.1 and the 2.2.15 kernel.

--Phil