From owner-freebsd-smp Mon Sep 23 11:50:15 2002 Delivered-To: freebsd-smp@freebsd.org Received: from mx1.FreeBSD.org (mx1.FreeBSD.org [216.136.204.125]) by hub.freebsd.org (Postfix) with ESMTP id 602EE37B401 for ; Mon, 23 Sep 2002 11:50:13 -0700 (PDT) Received: from vorbis.noc.easynet.net (vorbis.noc.easynet.net [195.40.1.254]) by mx1.FreeBSD.org (Postfix) with ESMTP id CF2F743E4A for ; Mon, 23 Sep 2002 11:50:12 -0700 (PDT) (envelope-from chrisy@vorbis.noc.easynet.net) Received: from chrisy by vorbis.noc.easynet.net with local (Exim 3.36 #1) id 17tYHX-000Bf2-00; Mon, 23 Sep 2002 19:50:07 +0100 Date: Mon, 23 Sep 2002 19:50:07 +0100 From: Chris Luke Cc: smp@freebsd.org Subject: Re: For those with P4 SMP problems.. Message-ID: <20020923185007.GA40537@flix.net> References: <3D8F48FB.F3DBE738@mindspring.com> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <3D8F48FB.F3DBE738@mindspring.com> User-Agent: Mutt/1.4i Organization: The Flirble Internet Exchange X-URL: http://www.flix.net/ X-FTP: ftp://ftp.flirble.org/ Sender: owner-freebsd-smp@FreeBSD.ORG Precedence: bulk List-ID: List-Archive: (Web Archive) List-Help: (List Instructions) List-Subscribe: List-Unsubscribe: X-Loop: FreeBSD.org Terry Lambert wrote (on Sep 23): > John Baldwin wrote: > > > The theory is that the BIOS has the corect information, but in the > > > wrong order, and FreeBSD cares about the order, but Linux and Windows Okay, here's something else to stir the pot, I have a Compaq Evo 6000 with 2x2GHz Xeon's. The mp_table is broken, as per Compaq standards. 2nd processor doesn't boot. No panics, but the cpu is reported by the kernel. But no "CPU #x launched!" message at the end of kernel boot. This is 4.6-STABLE as of the start of Septemeber (which was when I was last playing with it, I've left it/waited for someone else to find the problem - I was erring on the side of "It's Compaq, it's bound to be broken"). Anyway, the above issue seems similar-but-not- the-same to this thread. Something gets started, because start_ap is called, and just for the sake of it, I added some printf's to show this, and that it returned okay: number of APs: 1 start_ap called on logical cpu #1, physical cpu #6 AP 1 allegedly started okay FreeBSD/SMP: Multiprocessor motherboard cpu0 (BSP): apic id: 0, version: 0x00050014, at 0xfee00000 cpu1 (AP): apic id: 6, version: 0x00050014, at 0xfee00000 io0 (APIC): apic id: 8, version: 0x00178020, at 0xfec00000 The "called" message is early in start_ap(), the "allegedly" is after the call to start_ap(): --- CHECK_PRINT("trace"); /* show checkpoints */ + printf("AP %d allegedly started okay\n", x); /* record its version info */ cpu_apic_versions[x] = cpu_apic_versions[0]; --- If indeed the 2nd processor gets booted, it's almost as though it never sees the lock on the kernel released, so never does the stuff that prints that "launched!" message. I think I saw it boot up with the 2nd CPU once, but it was shortlived - the 2nd processor went away, almost as though I had imagined it (with "top", the only processes that were on CPU 1 were ones that had never run after startup - as they woke up, they all shifted to cpu 0 and new processes were always on cpu 0). mptable appears to agree with the output of acpidump, at least as far as processors and their apic ID's (and that hyperthreading is disabled). Can provide on request. Slight twist: This is my workstation in London, UK - I'm in Boston, US, so all my hacking has been with the cooperation of colleagues :) Also, it may be unrelated, but "clk" and "rtc" show twice as many irq's/s as I'd expect, when displayed with "systat -vmstat". 200Hz and 256Hz, respectively. The machine otherwise runs fine, with an SMP kernel - I'm sending this mail from it now. Just one cpu. Oh, another oddity, that may point to it being a Compaq-shonky issue - sometimes a warm boot will cause the BIOS to throw up "Processor initilization errors" on either 4, 2 or 1 processor(s), regardless of the setting of hyperthreading. The word "initialization" is spelled as typed too. Cold booting fixes it. (Given the choice, I'd have walked down the street and bought a m/b and bits and built the machine myself, but companies must have their procurement policies... grrr) Cheers for any light, Chris. -- == chrisy@flix.net To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-smp" in the body of the message