From owner-freebsd-scsi Tue May 15 1:10:21 2001 Delivered-To: freebsd-scsi@freebsd.org Received: from mass.dis.org (mass.dis.org [216.240.45.41]) by hub.freebsd.org (Postfix) with ESMTP id 30F5237B422 for ; Tue, 15 May 2001 01:10:03 -0700 (PDT) (envelope-from msmith@mass.dis.org) Received: from mass.dis.org (localhost [127.0.0.1]) by mass.dis.org (8.11.3/8.11.3) with ESMTP id f4ENTcO01438; Mon, 14 May 2001 16:29:39 -0700 (PDT) (envelope-from msmith@mass.dis.org) Message-Id: <200105142329.f4ENTcO01438@mass.dis.org> X-Mailer: exmh version 2.1.1 10/15/1999 To: Alexey Koptsevich Cc: scsi@FreeBSD.org Subject: Re: 3ware patch In-reply-to: Your message of "Mon, 14 May 2001 22:05:22 +0400." Mime-Version: 1.0 Content-Type: text/plain; charset=iso-8859-1 Mime-Version: 1.0 Content-Transfer-Encoding: quoted-printable Date: Mon, 14 May 2001 16:29:37 -0700 From: Mike Smith Sender: owner-freebsd-scsi@FreeBSD.ORG Precedence: bulk X-Loop: FreeBSD.org > = > Hello, > = > 3ware has released a patch for both firmware and software, Linux and > Windows, for RAID5 mode of their controllers. Is FreeBSD driver believe= d > to be free of errors? I've just exchanged email with my new technical contact at 3ware, and the= = problem as it's been explained to me will not affect FreeBSD. Here's the relevant sections from the mail describing the issue: ----8<---- RAID5 writes can result in corrupted parity.=A0 The data itself is not corrupted -- only the parity.=A0 The current Raid5 firmware performs _TWO= _ PCI accesses -- one for data, one for parity calculation.=A0 Thus if physical memory has changed before the second PCI access has completed, parity will be corrupted.=A0 This was a very bad oversight and has been corrected in the 7000, which only performs one PCI access. Linux ext2 FS performs a "fire and forget" on metadata.=A0 (Last write is= always the best write)=A0 These writes are asynchronous and never wait fo= r a completion. ----8<---- Since the FreeBSD block driver abstraction guarantees that data will not = be changed out from under the driver, there is no opportunity for this = limitation of the RAID implementation to cause problems. Note also that I've been informed that the 7000 series controllers will = not require the double-read behaviour, which should improve their RAID5 = performance considerably. -- = =2E.. every activity meets with opposition, everyone who acts has his rivals and unfortunately opponents also. But not because people want to be opponents, rather because the tasks and relationships force people to take different points of view. [Dr. Fritz Todt] V I C T O R Y N O T V E N G E A N C E To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-scsi" in the body of the message