From: Trond Myklebust Subject: Re: NFS4ERR_SYMLINK error Date: Mon, 09 Mar 2009 14:18:21 -0400 Message-ID: <1236622701.7234.34.camel@heimdal.trondhjem.org> References: <49AF9C28.8030603@cn.fujitsu.com> <49AFA4E4.50207@panasas.com> <20090306213234.GB1779@fieldses.org> <49B2C404.3060907@panasas.com> <20090309180643.GB9408@fieldses.org> Mime-Version: 1.0 Content-Type: text/plain Cc: Benny Halevy , Ni Wenjuan , linux-nfs@vger.kernel.org To: "J. Bruce Fields" Return-path: Received: from mail-out2.uio.no ([129.240.10.58]:34211 "EHLO mail-out2.uio.no" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751101AbZCISSd (ORCPT ); Mon, 9 Mar 2009 14:18:33 -0400 In-Reply-To: <20090309180643.GB9408@fieldses.org> Sender: linux-nfs-owner@vger.kernel.org List-ID: On Mon, 2009-03-09 at 14:06 -0400, J. Bruce Fields wrote: > > That said, I don't think a careful client implementation > > should ever get NFS4ERR_SYMLINK if it stats the directory it operates > > on before sending the link op (or lookup, create, rename, etc.) to make > > sure it is indeed a directory, right?. > > That sounds racy. The LINK op identifies the target directory by its filehandle, so I can't see that there can be any race: the server isn't supposed to be able to morph a directory into a symlink without changing its filehandle. Trond