From owner-freebsd-scsi Fri Dec 17 12:13:33 1999 Delivered-To: freebsd-scsi@freebsd.org Received: from implode.root.com (root.com [209.102.106.178]) by hub.freebsd.org (Postfix) with ESMTP id 6AC0414F81 for ; Fri, 17 Dec 1999 12:13:25 -0800 (PST) (envelope-from dg@implode.root.com) Received: from implode.root.com (localhost [127.0.0.1]) by implode.root.com (8.8.8/8.8.5) with ESMTP id MAA19446; Fri, 17 Dec 1999 12:11:03 -0800 (PST) Message-Id: <199912172011.MAA19446@implode.root.com> To: "Oliver Blasnik" Cc: freebsd-scsi@FreeBSD.ORG Subject: Re: Results of CRD5440-Checks In-reply-to: Your message of "Fri, 17 Dec 1999 12:43:06 +0100." <008101bf4883$e275cd70$da1940c2@omnilink.de> From: David Greenman Reply-To: dg@root.com Date: Fri, 17 Dec 1999 12:11:03 -0800 Sender: owner-freebsd-scsi@FreeBSD.ORG Precedence: bulk X-Loop: FreeBSD.org >The problem ONLY comes up after accessing both logical CRD-partitions (da0 and da1 in FreeBSD) in a high-load-state (40 dd's for one, 40 dd's for the other drive to fill up the command cache). If just accessing one partition, the system worked well. > >So, resulting fact is -> this controller does NOT work correctly, if tagged queuing enabled with more than one active logical partition. I switched off TQ at controller-level to let the kernel unchanged. Possibly this information should go to some type of hcl. > >I will post a full-featuerd report to crd-tech and hope to get a result. BTW, when you talk to CMD, ask them about another bug that I found with the CRD-5440: If you have a multiple of 640KB of dirty buffers in the write- back cache when a SCSI synchronize cache command is issued, the controller will lock up (crash). This can happen when you reboot the system (FreeBSD issues a synchronize cache command on last device close), but in practice is very unlikely - there isn't usually that much in the cache when the filesystem umount is done. To reproduce on a system with the CRD-5440 as da1 and not otherwise mounted, do: dd if=/dev/zero of=/dev/rda1 bs=64k count=10 ...will kill it every time, recoverable only with a power cycle. bs*count can be any multiple of 640KB. -DG David Greenman Co-founder/Principal Architect, The FreeBSD Project - http://www.freebsd.org Creator of high-performance Internet servers - http://www.terasolutions.com Pave the road of life with opportunities. To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-scsi" in the body of the message