From owner-freebsd-fs@FreeBSD.ORG Sun May 21 17:57:03 2006 Return-Path: X-Original-To: freebsd-fs@freebsd.org Delivered-To: freebsd-fs@freebsd.org Received: from mx1.FreeBSD.org (mx1.freebsd.org [216.136.204.125]) by hub.freebsd.org (Postfix) with ESMTP id DD01C16A8B3 for ; Sun, 21 May 2006 17:57:03 +0000 (UTC) (envelope-from kris@obsecurity.org) Received: from elvis.mu.org (elvis.mu.org [192.203.228.196]) by mx1.FreeBSD.org (Postfix) with ESMTP id 7EE4B43D5F for ; Sun, 21 May 2006 17:57:03 +0000 (GMT) (envelope-from kris@obsecurity.org) Received: from obsecurity.dyndns.org (elvis.mu.org [192.203.228.196]) by elvis.mu.org (Postfix) with ESMTP id 640DA1A4DAE for ; Sun, 21 May 2006 10:57:03 -0700 (PDT) Received: by obsecurity.dyndns.org (Postfix, from userid 1000) id CCDF452405; Sun, 21 May 2006 13:57:02 -0400 (EDT) Date: Sun, 21 May 2006 13:57:02 -0400 From: Kris Kennaway To: freebsd-fs@freebsd.org Message-ID: <20060521175702.GB21069@xor.obsecurity.org> References: <20060521132527.GA67386@hades.panopticon> Mime-Version: 1.0 Content-Type: multipart/signed; micalg=pgp-sha1; protocol="application/pgp-signature"; boundary="VrqPEDrXMn8OVzN4" Content-Disposition: inline In-Reply-To: <20060521132527.GA67386@hades.panopticon> User-Agent: Mutt/1.4.2.1i Subject: Re: NFS safety in 6.1-RELEASE X-BeenThere: freebsd-fs@freebsd.org X-Mailman-Version: 2.1.5 Precedence: list List-Id: Filesystems List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Sun, 21 May 2006 17:57:12 -0000 --VrqPEDrXMn8OVzN4 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline Content-Transfer-Encoding: quoted-printable On Sun, May 21, 2006 at 05:25:27PM +0400, Dmitry Marakasov wrote: > Hi! >=20 > In the TODO list for 6.1 RELEASE there was (and still is - > http://www.freebsd.org/releases/6.1R/todo.html) one entry that > bugs me: >=20 > Issue: NFS data corruption between two 7.0 machines > Status: In progress > Responsible: Mohan Srinivasan > Description: Running fsx between a 7.0 NFS client and server detects > data corruption. This problem can also be reproduced by using 6.1 > NFS server. The problem seems to be avoidable by turning off the > attribute cache on the NFS client. >=20 > So I wanted to ask what's status of NFS for release - was that issue > fixed and is it safe to use client and server NFS on 6.1 RELEASE? > If there's still data corruption possible, how do I turn off the > attribute cache? It's worth pointing out that the known problems either only manifest a) when running against specific servers (e.g. certain old Linux kernels, which probably means "server bug") b) under extremely high loads (my test suite uses 1000 concurrent filesystem I/O generators) Apart from the issues Scott mentioned, there is also a problem with writes via mmap() not being committed correctly, so subsequent reads do not see the correct data. This is not yet understood since it only occurs under condition b), so it's very hard to diagnose. For most users, these issues are unlikely to ever be encountered. Kris --VrqPEDrXMn8OVzN4 Content-Type: application/pgp-signature Content-Disposition: inline -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.3 (FreeBSD) iD8DBQFEcKnuWry0BWjoQKURAssIAJ4qkK7Xrp7/xDOzRIdUuj8wXYQItACeLUKu zkkBlEraD3CbiDq8wSU4hLc= =R4nF -----END PGP SIGNATURE----- --VrqPEDrXMn8OVzN4--