From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id ; Fri, 5 Jul 2002 08:32:43 -0400 Received: (majordomo@vger.kernel.org) by vger.kernel.org id ; Fri, 5 Jul 2002 08:32:42 -0400 Received: from pat.uio.no ([129.240.130.16]:4841 "EHLO pat.uio.no") by vger.kernel.org with ESMTP id convert rfc822-to-8bit; Fri, 5 Jul 2002 08:32:41 -0400 To: "Nils O." =?iso-8859-1?q?Sel=E5sdal?= Cc: linux-kernel@vger.kernel.org Subject: Re: df on a nfs mounted share vs local? References: <1025793209.10267.5.camel@space> From: Trond Myklebust Date: 05 Jul 2002 14:35:13 +0200 In-Reply-To: <1025793209.10267.5.camel@space> Message-ID: User-Agent: Gnus/5.0808 (Gnus v5.8.8) XEmacs/21.4 (Common Lisp) MIME-Version: 1.0 Content-Type: text/plain; charset=iso-8859-1 Content-Transfer-Encoding: 8BIT Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org >>>>> " " == Nils O > writes: > Just wondering, how come df reports root:~# df /mnt/export/ > Filesystem 1k-blocks Used Available Use% Mounted on /dev/hda5 > 2562252 383792 2178460 15% /mnt/export on the server, while on > a client that mounts /mnt/export over nfs: [root@space > download]# df /mnt/nfs/ Filesystem 1k-blocks Used Available > Use% Mounted on lfs:/mnt/export 2562256 383792 2178464 15% > /mnt/nfs > Just a few blocks diffrent, but I've seen much bigger.. also > seen +/- a few % on "Use" It is a rounding error. The block size on the NFS client is typically 4 or 8k, whereas the block size on the local filesystem is typically 512 bytes. On most UNIX implementations, the 'statvfs()' call supports two variables f_bsize and f_frsize, which allow you to distinguish between the two. Linux lacks kernel support for the latter variable. Cheers, Trond