The show-stopping (for some) TCP stalls on 2.1.1xx (starting around
2.1.109) at certain block sizes might still need looking into --
does anyone know if it has been sorted out, or where in the code
(I'm new to looking at the Linux networking code) one might start to
begin tests?
Reports of these problems continue. The difficulty manifests itself
here, and on at least one other reported machine, as TCP stalls on
large outgoing files after the kernel has been handling quite a lot
of traffic for a few hours -- a program to demonstrate this is
available at
http://www.billabong.demon.co.uk/downloads/socktest.tar.gz (this one
runs slowly even without running the kernel for some hours)
Once the file has stalled, TCP only works in little squirts...
making it run again requires that the interface be taken down and
re-started. Even then, the stall recurs when another large outgoing
file is sent.
An archive of posts so far on this topic is in
http://www.billabong.demon.co.uk/downloads/tcp-archive.gz
Reports have been noted so far from
bernd.paysan@gmx.de
Zlatko.Calusic@CARNet.hr
gmack@imag.net
F.K.W.van.de.Pol@inter.nl.net
linux@billabong.demon.co.uk
Any ideas? Tests we can run? Places in the code where we
should be looking? Any help will be gratefully received!
Yours,
John Hayward-Warburton
linux@billabong.demon.co.uk
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.altern.org/andrebalsa/doc/lkml-faq.html