From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (qmail 4747 invoked by alias); 28 Apr 2013 11:39:01 -0000 Mailing-List: contact ecos-discuss-help@ecos.sourceware.org; run by ezmlm Precedence: bulk List-Id: List-Subscribe: List-Archive: List-Post: List-Help: , Sender: ecos-discuss-owner@ecos.sourceware.org Received: (qmail 4737 invoked by uid 89); 28 Apr 2013 11:39:00 -0000 X-Spam-SWARE-Status: No, score=-1.0 required=5.0 tests=AWL,BAYES_50,RP_MATCHES_RCVD autolearn=ham version=3.3.1 Received: from netic.com (HELO netic.com) (208.131.130.247) by sourceware.org (qpsmtpd/0.84/v0.84-167-ge50287c) with ESMTP; Sun, 28 Apr 2013 11:39:00 +0000 Received: from PIII550 (60-242-91-36.static.tpgi.com.au [60.242.91.36]) by netic.com (8.13.1/8.13.1) with ESMTP id r3SBcutw011466 for ; Sun, 28 Apr 2013 05:38:57 -0600 From: "Laurie Gellatly" To: Date: Sun, 28 Apr 2013 11:39:00 -0000 Message-ID: <001201ce4404$f89d4f20$e9d7ed60$@netic.com> MIME-Version: 1.0 Content-Type: text/plain; charset="utf-8" Content-Transfer-Encoding: quoted-printable Subject: [ECOS] FreeBSD not always releasing ACK MBUFs X-SW-Source: 2013-04/txt/msg00030.txt.bz2 Hi All, I'm using eCosPro from pre eCos 3.0 to serve web pages with embedded images= , CSS and javascripts. Some pages are set to refresh their content and eventually the application = runs out of MBUFs and restarts. I can use my own web server or the CYGPKG_HTTPD server with the same result. I used the CYGPKG_HTTPD server on port 81 to monitor network stats. What I find is that the =E2=80=98HEADER=E2=80=99 type MBUFs grow and never = shrink, eventually running out of MBUFs completely. The amount of growth in each iteration varies. When I dump the data in the HEADER MBUFs (modified support.c to print these= short MBUFs) and compared them to the network traffic seen by wireshark I have confirmed that the HEA= DERs are correct ACK packets sent at the close of various HTTP get sequences. For some reason they are not being released. I also found that the number of timeout entries (NTIMEOUTS) in timeout.c (d= efault 8) is WAY below what is needed (about 70 being used). I'm thinking it's something to do with TIME_WAIT not closing the tcpcb when= it should. Has anyone seen this before or can suggest a way to troubleshoot this? Thanks ...Laurie:{) -- Before posting, please read the FAQ: http://ecos.sourceware.org/fom/ecos and search the list archive: http://ecos.sourceware.org/ml/ecos-discuss