From owner-freebsd-stable@FreeBSD.ORG Mon Mar 9 14:41:59 2009 Return-Path: Delivered-To: freebsd-stable@freebsd.org Received: from mx1.freebsd.org (mx1.freebsd.org [IPv6:2001:4f8:fff6::34]) by hub.freebsd.org (Postfix) with ESMTP id D267A1065672 for ; Mon, 9 Mar 2009 14:41:59 +0000 (UTC) (envelope-from varga.michal@gmail.com) Received: from mail-fx0-f158.google.com (mail-fx0-f158.google.com [209.85.220.158]) by mx1.freebsd.org (Postfix) with ESMTP id 406FB8FC19 for ; Mon, 9 Mar 2009 14:41:58 +0000 (UTC) (envelope-from varga.michal@gmail.com) Received: by fxm2 with SMTP id 2so1167717fxm.43 for ; Mon, 09 Mar 2009 07:41:57 -0700 (PDT) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gmail.com; s=gamma; h=domainkey-signature:mime-version:received:date:message-id:subject :from:to:content-type:content-transfer-encoding; bh=5hHoCFYYLvh2SIqaWOhLVrFxrj7puAGxlIia1U6bYoU=; b=Jrkyd+BvDul9SsZM5nfvA4zyfkz9pbeCUCBehIAGTlFYIZzx7+X5e/CVrwLoMjBxBE n5fIKPh7XOHBKOX+Ip1FmeYYZyUWT///FNLzH+7NkLoKbODQKQeIHiwADznw4aU8MMFW 2HfibeJLo8SLYspOR0WmeCClWxIqWG39mN4qA= DomainKey-Signature: a=rsa-sha1; c=nofws; d=gmail.com; s=gamma; h=mime-version:date:message-id:subject:from:to:content-type :content-transfer-encoding; b=BvT6v5si92dQEAKuTKHn2fZJJBiPNFMDGc78F5U7kGO9piAZL4YQeuXeB+BufuCvgc UH0BPGInWwDZwc16HiR4c7hHn7UwCthJJfW/VpiO6GKZ0Dtm/9mJlTLYOu1388DJEucU 3QIBRqMdRisWveaDSF0pT1+k7hxqoyfXsD0t4= MIME-Version: 1.0 Received: by 10.223.119.198 with SMTP id a6mr4530290far.42.1236608165611; Mon, 09 Mar 2009 07:16:05 -0700 (PDT) Date: Mon, 9 Mar 2009 15:16:05 +0100 Message-ID: <3f1fd1ea0903090716k79ca91bemb21aff9db0073e6e@mail.gmail.com> From: Michal Varga To: freebsd-stable@freebsd.org Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 7bit Subject: SCHED_ULE + SMP Phenom freeze X-BeenThere: freebsd-stable@freebsd.org X-Mailman-Version: 2.1.5 Precedence: list List-Id: Production branch of FreeBSD source code List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Mon, 09 Mar 2009 14:42:00 -0000 Guys, I'd like to point out a still present (rather serious I guess) problem with SCHED_ULE, possibly triggered when Phenom X3 CPUs are used. First time I encountered this problem in early January, it took me quite a while to narrow the freeze down to ULE with SMP support, well, after eliminating every other possibility one by one. After that, it wasn't that hard to find that I'm not the only one seeing this: http://lists.freebsd.org/pipermail/freebsd-stable/2008-August/044167.html http://lists.freebsd.org/pipermail/freebsd-stable/2008-October/045714.html (continued) http://lists.freebsd.org/pipermail/freebsd-amd64/2008-December/011722.html (there were few more with the same symptoms back then, but sadly I can't figure out the right keywords to google them at the moment) Basically, the system can't get past the boot stage with ULE + SMP support in kernel and hangs after "SMP: AP CPU #X Launched!" (see the linked October thread for details). But it gets better - there are absolutely no problems with either uniprocessor SCHED_ULE, nor SCHED_4BSD (both UP and SMP). No interrupt storms, no suspicious behaviour, no crashes, everything acts rock solid. I've been running this system with SCHED_ULE minus SMP support for more than a month under heavy load, tested SCHED_4BSD minus SMP for a brief moment, and then another month with SCHED_4BSD + SMP. Everything works perfectly and also all three cores behave correctly with 4BSD + SMP. It only gets messy when ULE + SMP combo is used. I was also wondering why only Phenom X3 people report this problem, but this may be just a coincidence, who knows. Still, I'm also one of them: FreeBSD 7.1-STABLE #0: Sat Mar 7 21:37:27 CET 2009 root@xenon:/usr/obj/usr/src/sys/KERNEL Timecounter "i8254" frequency 1193182 Hz quality 0 CPU: AMD Phenom(tm) 8450 Triple-Core Processor (2109.74-MHz 686-class CPU) Origin = "AuthenticAMD" Id = 0x100f23 Stepping = 3 Features=0x178bfbff Features2=0x802009 AMD Features=0xee500800 AMD Features2=0x7ff,,,Prefetch,,> TSC: P-state invariant Cores per package: 3 Same as with other reports, with SCHED_ULE + SMP, as soon as I hit the "SMP: AP CPU Launched" stage, the system is dead. m.