From owner-freebsd-stable Tue Jul 16 6:55:55 2002 Delivered-To: freebsd-stable@freebsd.org Received: from mx1.FreeBSD.org (mx1.FreeBSD.org [216.136.204.125]) by hub.freebsd.org (Postfix) with ESMTP id 9107837B426 for ; Tue, 16 Jul 2002 06:55:36 -0700 (PDT) Received: from hcshh.hcs.de (hcshh.hcs.de [194.123.40.1]) by mx1.FreeBSD.org (Postfix) with ESMTP id C676644020 for ; Tue, 16 Jul 2002 06:51:27 -0700 (PDT) (envelope-from hm@hcs.de) Received: from localhost (localhost.hcs.de [127.0.0.1]) by hcshh.hcs.de (Postfix) with ESMTP id 1530E15580 for ; Tue, 16 Jul 2002 15:32:36 +0200 (CEST) Received: from hcswork.hcs.de (hcswork.hcs.de [172.24.124.5]) by hcshh.hcs.de (Postfix) with ESMTP id 4247A155C7 for ; Tue, 16 Jul 2002 15:32:34 +0200 (CEST) Received: by hcswork.hcs.de (Postfix, from userid 200) id 13299566; Tue, 16 Jul 2002 15:32:33 +0200 (METDST) Subject: "Hanging" TCP connections over ISDN To: FreeBSD-Stable List Date: Tue, 16 Jul 2002 15:32:33 +0200 (METDST) Reply-To: hm@hcs.de Organization: HCS Hanseatischer Computerservice GmbH X-Mailer: ELM [version 2.4ME+ PL84 (25)] MIME-Version: 1.0 Content-Transfer-Encoding: 7bit Content-Type: text/plain; charset=US-ASCII Message-Id: <20020716133233.13299566@hcswork.hcs.de> From: hm@hcs.de (Hellmuth Michaelis) X-Virus-Scanned-HCS: by AMaViS perl-11 Sender: owner-freebsd-stable@FreeBSD.ORG Precedence: bulk List-ID: List-Archive: (Web Archive) List-Help: (List Instructions) List-Subscribe: List-Unsubscribe: X-Loop: FreeBSD.ORG Hi, since upgrading two ISDN-connected systems from 4.5 RC1 to 4.6 TCP connections between this two systems "hang" from time to time. When both ran under 4.5, interactive connections (mostly ssh and scp) run smoothly, no "hangs" or interruptions were noticed. The two systems ran (nearly) same OS versions using i4b and userland ppp(8). When 4.6 arrived, i decided to test the RC's and since then both systems ran RC1, RC2 and now a stable from the first week of July. Since the first 4.6 i began noticing "hangs" when using interactive programs (elm, editors, etc.) which occur now and then sometimes more often, sometimes less, sometimes in intervalls of 3 or 5 seconds - i cant see a regularity in it. With scp, data flows, gets slower, stalls, flows, stalls, flows etc. all in all it the transfer rate much lower than the expected 7.5 Kb/s: about 1.5 to 3 Kb/s. First i traced the ISDN traffic: all seems well, no anomalies, just the B-channel traffic stopped sometines, see above. Then i saw a mail from phk "Is newreno still sick ?": i disabled newreno on all machines and it got much better: where the "hangs" lasted several seconds, the now lasted just one or two seconds and i "felt" they occured much less. Still, it is boring to work over such a link. Last weekend i instrumented the i4b rbch-driver, but as hard as i tried, i found nothing wrong. Then i reduced the mtu and mru (1500 by default) for ppp(8) and the smaller i made it, the smoother the connection became. Still "hangs" but the frequency got lower and the hangs were much shorter - nearly gone. Now today i replaced the combination i4b-rbch/userland-ppp with the i4b isp/sppp combination and i get the same hangs as with the default rbch/ppp setup. I'm a bit stuck here. It still might be a bug in the ISDN subsystem, but nothing important has changed between 4.5 and 4.6. Two totally different ISDN interfaces to the network stack show the same behaviour. So it might not be a bug in the ISDN subsystem. Using several different parameters in the networking code leads to a changed behaviour. So it might be a bug in the network subsystem ;-) As i said, i'm a bit stuck here. Perhaps someone has an idea or a hint for me where to have a more detailed look or what knobs to use to fix this problem. hellmuth -- Hellmuth Michaelis Tel +49 40 55 97 47-70 HCS Hanseatischer Computerservice GmbH Fax +49 40 55 97 47-77 Oldesloer Strasse 97-99 Mail hm [at] hcs.de D-22457 Hamburg WWW http://www.hcs.de To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-stable" in the body of the message