From owner-freebsd-stable@FreeBSD.ORG Tue Dec 20 02:10:00 2005 Return-Path: X-Original-To: freebsd-stable@freebsd.org Delivered-To: freebsd-stable@freebsd.org Received: from mx1.FreeBSD.org (mx1.freebsd.org [216.136.204.125]) by hub.freebsd.org (Postfix) with ESMTP id E15E316A41F for ; Tue, 20 Dec 2005 02:10:00 +0000 (GMT) (envelope-from mandrews@bit0.com) Received: from mindcrime.bit0.com (bit0.com [216.24.42.169]) by mx1.FreeBSD.org (Postfix) with ESMTP id 685C943D72 for ; Tue, 20 Dec 2005 02:09:53 +0000 (GMT) (envelope-from mandrews@bit0.com) Received: from localhost (localhost.bit0.com [127.0.0.1]) by mindcrime.bit0.com (Postfix) with ESMTP id 5C40C730002 for ; Mon, 19 Dec 2005 21:09:52 -0500 (EST) Received: from mindcrime.bit0.com ([127.0.0.1]) by localhost (mindcrime.bit0.com [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id 02717-01 for ; Mon, 19 Dec 2005 21:09:51 -0500 (EST) Received: from localhost (localhost.bit0.com [127.0.0.1]) by mindcrime.bit0.com (Postfix) with ESMTP for ; Mon, 19 Dec 2005 21:09:51 -0500 (EST) Date: Mon, 19 Dec 2005 21:09:51 -0500 (EST) From: Mike Andrews To: freebsd-stable@freebsd.org Message-ID: <20051219210527.G2707@bit0.com> MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII; format=flowed X-Virus-Scanned: by amavisd-new at bit0.com Subject: "Request Requeued/Retrying Command" with twa card on FreeBSD 6.0 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: Tue, 20 Dec 2005 02:10:01 -0000 I've got a weird but apparently minor issue with a 3Ware 9500S-4LP in a FreeBSD 6.0-RELEASE system... When the system is run with boot.verbose="YES" in /boot/loader.conf, I get these messages on the console: (da0:twa0:0:0:0): Request Requeued (da0:twa0:0:0:0): Retrying Command (da0:twa0:0:0:0): Request Requeued (da0:twa0:0:0:0): Retrying Command (repeat a few dozen times each time) ...during some heavy disk activity, like, for example, qpopper rewriting a 50 MB mailbox. (I know qpopper's a bit of a slug.) Not all heavy disk IO triggers it, though; I can do something like "dd if=/dev/zero of=/tmp/testfile bs=10485760 count=128" followed by "dd if=/tmp/testfile of=/dev/null bs=10485760" and have it complete without a single SCSI timeout message. Turn boot.verbose="NO" off and the messages go away, but IO still lags badly. "camcontrol tags" says it has 254 tag openings, which seems like a bit much. Dropping down to 16 seems to make the messages go away. Do the twa cards even do TCQ between the card and the OS? From what I understand, they won't do NCQ between the card and the SATA disks, even if you have something like Barracudas that support it -- you need the new 3Ware 9550 card for that. If they don't, why does it advertise 254 tags, and if it does, why does it seem to choke on the default setting of 254? Other than laggy IO (shell prompts hang, MP3's skip for 5 - 30 seconds) there aren't any actual IO errors or anything else wrong, which for this particular system is just minor annoyance instead of serious problem. 5.4-RELEASE did the same thing. I have an apparent workaround, I'm just trying to sanity-check it and its necessity. This is an Asus P4C800-E Deluxe board, so I'm running the twa card in a 32-bit slot. The system is otherwise totally idle when this happens, no interrupt storms, USB and Firewire disabled in the BIOS so we don't have IRQs shared with Giant-locked drivers... dmesg, vmstat -i, other stuff available on request if needed. Kernel config is the stock 'SMP' one. Mike Andrews * mandrews@bit0.com * http://www.bit0.com It's not news, it's Fark.com. Carpe cavy!