From owner-freebsd-smp Thu Mar 6 3:10:30 2003 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 9B00137B401 for ; Thu, 6 Mar 2003 03:10:28 -0800 (PST) Received: from c-24-130-253-104.we.client2.attbi.com (c-24-130-253-104.we.client2.attbi.com [24.130.253.104]) by mx1.FreeBSD.org (Postfix) with ESMTP id 18F9843F85 for ; Thu, 6 Mar 2003 03:10:28 -0800 (PST) (envelope-from obrien@c-24-130-253-104.we.client2.attbi.com) Received: from localhost (localhost [127.0.0.1]) by c-24-130-253-104.we.client2.attbi.com (8.11.1/8.11.1) with ESMTP id h26B9nd00841 for ; Thu, 6 Mar 2003 03:09:49 -0800 (PST) (envelope-from obrien@c-24-130-253-104.we.client2.attbi.com) Message-Id: <200303061109.h26B9nd00841@c-24-130-253-104.we.client2.attbi.com> X-Mailer: exmh version 2.2 06/23/2000 with version: MH 6.8.4 #1[UCI] To: freebsd-smp@freebsd.org Subject: SMP hang on boot in RELENG_4_7 on Iwill DP533 board Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Date: Thu, 06 Mar 2003 03:09:48 -0800 From: "Mike O'Brien" 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 I'm running dual Xeon processors on an Iwill DP533 board. This board has an E7505 MCH chipset from Intel, using the ICH4 I/O hub and an 82870P2 PCI-X controller. This configuration does the same thing that John Stockdale reported on his Tyan Thunder i7505 board: it runs fine with a uniprocessor kernel, but attempting to boot an SMP kernel results in a hang at the point where the kernel has just printed "APIC_IO: Testing 8254 interrupt delivery." Rather than muck with my BIOS as he did, I mucked with the kernel. This message is printed by clock.c, just before executing a "while" loop waiting for some I/O register action. That action never happens. Not figuring it would work, but figuring it'd at least be instructive, I inserted a count-backward-from-a-million escape hatch so the loop would run for a while, then exit. The resulting kernel did boot, and did turn on the second processor, and seemed to run OK until I tried to do a "make installworld". I got a ton of disk read and write timeout errors, probably because the APIC_IO wasn't, in fact, set up correctly. My understanding is that a good part of these problems are that the developers don't have access to the hardware needed to test stuff out. Any SMP developer who wants to work with me on testing fixes to this problem, send me private email, please, and let's see if we can work something out. I for dang sure don't want to be limited to a single processor. Suggestions welcome. I don't know, for example, if the APIC setup in 5.0 has been improved to the point where I should try running it. I was avoiding it, but I could give it a try if there's a chance it'll work. Comments? Mike O'Brien To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-smp" in the body of the message