From owner-freebsd-emulation@FreeBSD.ORG Sat Jan 26 20:39:30 2008 Return-Path: Delivered-To: freebsd-emulation@freebsd.org Received: from mx1.freebsd.org (mx1.freebsd.org [IPv6:2001:4f8:fff6::34]) by hub.freebsd.org (Postfix) with ESMTP id 4F6A616A41B for ; Sat, 26 Jan 2008 20:39:30 +0000 (UTC) (envelope-from mathom@pifpaf.pl) Received: from main.pifpaf.pl (main.pifpaf.pl [193.42.218.2]) by mx1.freebsd.org (Postfix) with ESMTP id 77AAF13C46A for ; Sat, 26 Jan 2008 20:39:27 +0000 (UTC) (envelope-from mathom@pifpaf.pl) Received: from mathom.promienista (esk32.neoplus.adsl.tpnet.pl [83.20.130.32]) (authenticated bits=0) by Mail Exchanger main.pifpaf.pl with ESMTP id m0QKdNrc017709; Sat, 26 Jan 2008 21:39:23 +0100 (CET) (envelope-from mathom@pifpaf.pl) Date: Sat, 26 Jan 2008 21:39:21 +0100 From: Tomasz Formanowski X-Mailer: The Bat! (v3.5.26) Professional X-Priority: 3 (Normal) Message-ID: <6340678.20080126213921@pifpaf.pl> To: Alexander Leidinger In-Reply-To: <20080126205431.682b9401@deskjail> References: <1096069283.20080122130241@pifpaf.pl> <20080122151941.GA93141@freebsd.org> <157462854.20080126130147@pifpaf.pl> <20080126205431.682b9401@deskjail> MIME-Version: 1.0 Content-Type: multipart/mixed; boundary="----------331B88E2B1F8CF8" X-Spam-Flag: NO X-Spam-Status: No, score=-102.3 required=5.0 tests=AWL,BAYES_00, FORGED_RCVD_HELO,USER_IN_WHITELIST autolearn=ham version=3.1.7 X-Spam-Checker-Version: SpamAssassin 3.1.7 (2006-10-05) on mail.pifpaf.pl X-Virus-Scanned: ClamAV version 0.90.1, clamav-milter version 0.90.1 on main.pifpaf.pl X-Virus-Status: Clean X-Content-Filtered-By: Mailman/MimeDel 2.1.5 Cc: freebsd-emulation@freebsd.org Subject: Re[2]: getdents() X-BeenThere: freebsd-emulation@freebsd.org X-Mailman-Version: 2.1.5 Precedence: list Reply-To: Tomasz Formanowski List-Id: Development of Emulators of other operating systems List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Sat, 26 Jan 2008 20:39:30 -0000 ------------331B88E2B1F8CF8 Content-Type: text/plain; charset=Windows-1250 Content-Transfer-Encoding: 8bit Witaj Alexander, Dnia Saturday, January 26, 2008, 8:54:31 PM, napisałeś : AL> Quoting Tomasz Formanowski (Sat, 26 Jan 2008 13:01:47 +0100): >> RD> On Tue, Jan 22, 2008 at 01:02:41PM +0100, Tomasz Formanowski wrote: >> >> Hi, >> >> >> >> Is there any hope that this: >> >> http://www.freebsd.org/cgi/query-pr.cgi?pr=117010 >> >> >> >> will be fixed soon ? >> >> >> >> Linux emulation is unusable now for newer software, like game servers >> >> etc. >> >> RD> in private mails we agreed that tomasz/his-company and me will try >> RD> to cooperate on this on... >> >> RD> hopefully this will fruit some results >> >> >> We've looked through the code and decided to give linux a try. We've >> been using freebsd for 5 years in production environment and it's a >> good system. So we wont be fixing this bugs, sorry. If it were only >> this one bug, than why not, but ... >> >> There were problems with: SMP support, which we hoped would be fixed >> in 7.x releases. And maybe it were, but linuxolator is a mess there. AL> Uhm. It works for a lot of people and we even run the linux regression AL> tests form the linux test project AL> (http://wiki.freebsd.org/linux-kernel/ltp), so if you have problems it AL> would be nice if you could provide some more info than just "a mess". AL> Could you please give instructions how to reproduce the problems you AL> have? >> Bottomline is, that not only 2.6 kernel support is incomplete, there AL> Which version did you use for compat.linux.osrelease? Anything apart AL> from 2.4.2 and 2.6.16 is not supported ATM. >> is also faulty 2.4 support. Apart from getdents() problems (which >> probably is really a deeper problem - looks like programs crash not on >> getdents() but on closedir(). just as something had freed the >> descriptor), there seems to be a problem with fork() event in 2.4 >> kernel and fc4 libs. Programs like team speak 2 and unreal-server >> crash when trying to fork. AL> I remember reports from users of team speak that it works for them. AL> What is the osrelease value for the tests you did? Which linux base did AL> you use? Is this an install from scratch, or did you test on a system AL> which has seen updates of the linux ports? If it is the later, are you AL> willing to test again with a clean /compat/linux/ (no linux ports AL> installed) and a reinstall of the linux ports? Does the system survive AL> several buildworlds in a row (to rule out hardware problems)? Which AL> FreeBSD version did you test? TS2 and UT works just fine on bsd <=6.2. I'm talking about 7.x (with default linux_base and 2.4 kernel). There wasn't many problems with older bsd's and linuxolator 2.4. Unfortunately some new games require either kernel 2.6 or glibc 2.6. We've tested the following software on *bsd 7 with fc4*: - Team Speak 2 voice server (/usr/ports/audio/teamspeak_server) Problem: This server dies on startup (core dump). Interestingly it runs fine on 7-beta3 with fc7/kernel 2.6. - Unreal Tournament 99 server (/usr/ports/games/ut-server, we didn't try with ut2003, 2004 and ut3, but i'm pretty sure there will be the same problem there) Problem: The server advertises itself with game master servers. There usualy are 3 default master servers set up. In this configuration server crashes on fork (). It works fine with one master server specified. - HLDS/SRCDS (Valve steam engine) on fc6/fc7, both 7-beta and 7-rc1 Problem: getdents() random crashes. What's interesting, this problem seems to depend on files order in the directory. Try moving some files from http://www.freebsd.org/cgi/query-pr.cgi?pr=117010 test programm from and to temp directory to change files order. getdents() works with some files order, and doesn't work with default one (ie. alphabetical order). What's strange, even with core dump, all files are read fine. Program crashes with double free *after* getdents invokation, maybe on closedir. There may be something with frees the descriptor, maybe sth in glibc. Problem 2: we are also struggling with hdsl/srcds (counter-strike and counter-strike:source) random core dumps for some time now. This problem is hard to replicate and trace (we can't trace all production servers), and seems to be nonexistent on linux. This applies to bsd 4.x, 5.x and 6.x. I've pointed ported distros of this software to give you easier environment for problem reproduction. I've attached kdumps from both with sample server.ini config for ut (put it in System directory and run server with utcmdline.sh from server root dir) pozdrawiam Tomek "Mathom" Formanowski ------------331B88E2B1F8CF8--