Return-Path: Delivered-To: apmail-httpd-dev-archive@httpd.apache.org Received: (qmail 91836 invoked by uid 500); 4 Dec 2002 23:38:05 -0000 Mailing-List: contact dev-help@httpd.apache.org; run by ezmlm Precedence: bulk Reply-To: dev@httpd.apache.org list-help: list-unsubscribe: list-post: Delivered-To: mailing list dev@httpd.apache.org Received: (qmail 91823 invoked from network); 4 Dec 2002 23:38:04 -0000 Date: Wed, 4 Dec 2002 23:38:06 +0000 From: =?iso-8859-1?Q?Colm_MacC=E1rthaigh?= To: Jeff Trawick , dev@httpd.apache.org, dev@apr.apache.org Subject: Re: Linux + TCP_CORK + IPv6 = Broken [PATCH] Message-ID: <20021204233806.A12474@prodigy.Redbrick.DCU.IE> References: <20021203213808.GA2974@byron.heanet.ie> <004d01c29b1c$002046a0$210d640a@unfix.org> <20021204092734.A17772@prodigy.Redbrick.DCU.IE> <20021204122549.A28623@prodigy.Redbrick.DCU.IE> <20021204123039.A14108@prodigy.Redbrick.DCU.IE> <20021204230521.GA619@redhat.com> Mime-Version: 1.0 Content-Type: text/plain; charset=iso-8859-1 Content-Disposition: inline Content-Transfer-Encoding: 8bit User-Agent: Mutt/1.2.5i In-Reply-To: <20021204230521.GA619@redhat.com>; from jorton@redhat.com on Wed, Dec 04, 2002 at 11:05:21PM +0000 X-Spam-Rating: daedalus.apache.org 1.6.2 0/1000/N On Wed, Dec 04, 2002 at 11:05:21PM +0000, Joe Orton wrote: > On Wed, Dec 04, 2002 at 08:21:36AM -0500, Jeff Trawick wrote: > > Colm MacC�rthaigh writes: > > > > > On Wed, Dec 04, 2002 at 12:25:49PM +0000, Colm MacC�rthaigh wrote: > > > > My tests (and patch) were based on apr and apr-util from CVS , > > > > with the 2.0.43 codebase, because CVS seems broken right now. > > > > > > stupid pre-test patch, here's the real one: > > > > patch looks basically right to me... probably is exactly right... > > unless somebody beats me to it I'll play with it and commit it... nag > > me directly if it isn't committed in 36 hours or so :) > > It would seem prudent to confirm this is a kernel bug before adding a > workaround to APR. APR's test/sendfile works fine here over IPv6 > between two 2.4.18-based kernels (RHL 8.0), and is doing basically the > same thing as httpd, albeit without the TCP_NODELAY - maybe that makes a > difference. TCP_CORK and TCP_NODELAY are mutually exclusive anyway, so it wouldnt make any difference :) As for it's status as a kernel bug, I'm more and more certain that it is, I don't think the kernel is recycling a file descriptor properly (this is purely based on kdb, I havnt started trawling source yet). I'm more familiar with the apache source (grep -c colmmacc CHANGES ;) , it's a while since I played with the Linux kernel internals. > Colm, can you try running the apr/test/sendfile binary with your > machines? It was one of the first things I tried when I was debugging, unfortunately, it doesnt seem to support IPv6 from the client side at all, I get: colmmacc@byron:~/apr/test$ ./sendfile client blocking orion.ipv6.heanet.ie Creating a test file... apr_sockaddr_info_get()->670005/No address associated with hostname colmmacc@byron:~/apr/test$ host -t AAAA orion.ipv6.heanet.ie orion.ipv6.heanet.ie AAAA 2001:770:18:2:206:5BFF:FE8D:2402 colmmacc@byron:~/apr/test$ ./sendfile client blocking 2001:770:18:2:206:5BFF:FE8D:2402 Creating a test file... apr_sockaddr_info_get()->670009/Address family for hostname not supported Although, the server process does seem to be listening on IPv6. tcp 0 0 :::8021 :::* LISTEN -- colmmacc@redbrick.dcu.ie PubKey: colmmacc+pgp@redbrick.dcu.ie Web: http://devnull.redbrick.dcu.ie/