From owner-freebsd-hackers Thu Nov 22 9:36: 3 2001 Delivered-To: freebsd-hackers@freebsd.org Received: from iguana.aciri.org (iguana.aciri.org [192.150.187.36]) by hub.freebsd.org (Postfix) with ESMTP id 95EE137B405 for ; Thu, 22 Nov 2001 09:35:59 -0800 (PST) Received: (from rizzo@localhost) by iguana.aciri.org (8.11.3/8.11.1) id fAMHVvY47257; Thu, 22 Nov 2001 09:31:57 -0800 (PST) (envelope-from rizzo) Date: Thu, 22 Nov 2001 09:31:57 -0800 From: Luigi Rizzo To: rick norman Cc: freebsd-hackers@freebsd.org Subject: Re: dummynet "pipe show" problem Message-ID: <20011122093157.B47061@iguana.aciri.org> References: <3BDEF5C5.F5D604C2@lmco.com> <20011030113914.H9665@iguana.aciri.org> <3BFC3299.E7DBA702@ikaika.com> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <3BFC3299.E7DBA702@ikaika.com> User-Agent: Mutt/1.3.23i Sender: owner-freebsd-hackers@FreeBSD.ORG Precedence: bulk List-ID: List-Archive: (Web Archive) List-Help: (List Instructions) List-Subscribe: List-Unsubscribe: X-Loop: FreeBSD.ORG On Wed, Nov 21, 2001 at 03:02:49PM -0800, rick norman wrote: > Hi, > I'm still running into lots of problems with this on 4.3. Is it neccessary > to introduce a delay in the pipe ? Would a delay of zero work ? > Has this been fixed in 4.4 ? Thinking about it, I believe it might be a problem with the routing code, not with dummynet, and it has to do with the fact that you delete and reassign the address to the loopback interface. I am almost sure that 4.4 would cure the problem, because that problem was fixed before 4.4. cheers luigi > Rick > > Luigi Rizzo wrote: > > > On Tue, Oct 30, 2001 at 10:47:33AM -0800, rick norman wrote: > > > Hi, > > > I have enclosed a short piece of code that seems to > > > reproduce the problem after 5 to 10 minutes. > > > I am running 4.3 freebsd off the release cd's. > > > Pretty much the generic kernel except for the > > > addition of dummynet and ipfw. My platform > > > is a dell dimension 4100 with 1ghz p3, though > > > I doubt that is relevant. To reproduce the problem, > > > compile and run the enclosed code in one window. > > > In another window su to root and run "ping -s 1024 -f 127.0.42.1". > > > As you will see, the code reports the gowing byte count as one > > > would expect. Walk away for 5 to 10 minutes and when you come > > > back you should see the state I'm talking about. The flood ping stream > > > > It _might_ be a locking problem due to the frequent > > reconfigurations of the pipe -- i think there was some > > fix of this kind related to ipfw commands between 4.3 and 4.4. > > I will see if i can reproduce the problem locally (but i have > > 4.4). > > > > You are using dummynet in a very peculiar way: > > more precisely, your pipe does not introduce any delay, > > and this excludes the scheduler which is part of dummynet. > > Also you are using the same pipe for icmp requests > > and replies, which are generated within the kernel, so > > the behaviour should be quite deterministic in this > > respect. > > > > thanks for the report > > > > luigi > > > > > Thanks, > > > Rick > > > > > > ----------------------cut-------------------------------------------- > > > > > > main() { > > > int i; > > > > > > restart: > > > system("ifconfig lo0 -alias 127.0.42.1"); > > > system("ifconfig lo0 alias 127.0.42.1 netmask 255.255.255.0"); > > > system("ipfw -f flush"); > > > system("ipfw pipe 1 delete"); > > > system("ipfw add pipe 1 ip from 127.0.42.0/24 to any in"); > > > > > > for(i=0;i<5;i++) { > > > system("ipfw pipe 1 config queue 2048Bytes"); > > > sleep(1); > > > system("ipfw pipe 1 show"); > > > } > > > > > > goto restart; > > > } > > > > > > --------------------cut-------------------------------------- > > > > > > > > > > > > To Unsubscribe: send mail to majordomo@FreeBSD.org > > > with "unsubscribe freebsd-hackers" in the body of the message > > > > To Unsubscribe: send mail to majordomo@FreeBSD.org > > with "unsubscribe freebsd-hackers" in the body of the message > To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-hackers" in the body of the message