From: Jonathan Woithe Subject: Re: Clients not displaying NFS quotas Date: Mon, 7 May 2007 09:50:23 +0930 (CST) Message-ID: <200705070020.l470KNxD004058@turbo.physics.adelaide.edu.au> References: <1178496403.6493.18.camel@heimdal.trondhjem.org> Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Cc: nfs@lists.sourceforge.net, Jonathan Woithe To: trond.myklebust@fys.uio.no (Trond Myklebust) Return-path: Received: from sc8-sf-mx2-b.sourceforge.net ([10.3.1.92] helo=mail.sourceforge.net) by sc8-sf-list2-new.sourceforge.net with esmtp (Exim 4.43) id 1Hkqy0-0003dl-HY for nfs@lists.sourceforge.net; Sun, 06 May 2007 17:20:40 -0700 Received: from adelphi.physics.adelaide.edu.au ([129.127.102.1]) by mail.sourceforge.net with esmtps (TLSv1:AES256-SHA:256) (Exim 4.44) id 1Hkqy1-0008Eg-Jk for nfs@lists.sourceforge.net; Sun, 06 May 2007 17:20:43 -0700 In-Reply-To: <1178496403.6493.18.camel@heimdal.trondhjem.org> from "Trond Myklebust" at May 06, 2007 08:06:43 PM List-Id: "Discussion of NFS under Linux development, interoperability, and testing." List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Sender: nfs-bounces@lists.sourceforge.net Errors-To: nfs-bounces@lists.sourceforge.net > On Mon, 2007-05-07 at 08:57 +0930, Jonathan Woithe wrote: > > > On Thu, 2007-05-03 at 11:43 +0930, Jonathan Woithe wrote: > > > > We have a Slackware 9.1 box (with numerous software upgrades) exporting a > > > > directory to various NFS clients. While running a 2.4.x kernel this all > > > > worked fine - users on clients could use "quota" to display their quota > > > > information. When the system was moved to a 2.6.19.2 kernel a few months > > > > ago the clients could no longer display quota information. The server > > > > still knew and enforced the quotas, it's just that from the client's > > > > point of view there were no quotas. > > > > > > > > The system is running the following software versions: > > > > > > > > * kernel 2.6.19.2 > > > > * nfs-utils 1.0.10 > > > > * util-linux 2.12 > > > > * portmap 5.0 > > > > > > > > The filesystem concerned is reiserfs3. The kernel has reiserfs quota > > > > support enabled. > > > > > > > > The relevant partition is mounted on the server using an fstab entry of > > > > > > > > /dev/... /home/foobar reiserfs usrquota > > > > > > > > All clients mount the volume via the automounter with the "hard,intr,bg" > > > > mount options. > > > > > > > > Users cannot log into the NFS server. > > > > > > > > We have clients running various versions of redhat, fedora and slackware > > > > with different kernels. In all cases things worked with the 2.4.x > > > > kernel and don't now. It therefore seems to be a server-related issue. > > > > > > > > Does anyone have any idea as to why client access to quota information > > > > isn't working in this scenario? > > > > > > The NFS protocol doesn't support quotas, so those have to be transmitted > > > using a sideband protocol. Check that you have the 'quotad' daemon > > > running on your server, and that it is accessible from the clients. > > > > Yes, quotad is running on the server: > > > > > rpcinfo -p > > : > > 100011 1 udp 998 rquotad > > 100011 2 udp 998 rquotad > > > > As I mentioned, everything works just fine if a 2.4.x kernel is running, > > which presumedly clears any access issues (unless of course the 2.6.x kernel > > is doing something odd). The moment 2.6.x is booted (where x = 19.2 > > currently) the clients can no longer see quota information. > > > > Is there a way to explicitly test whether clients can in fact connect to > > rquotad? > > > > I have noticed that rquotad is only listening on UDP, rather than on both > > TCP and UDP as I think it did under 2.4.x. Is this likely to be the > > problem? If so, what would cause rquotad to do this under 2.6.x given an > > unchanged configuration from 2.4.x? > > > > Here are the relevant kernel configuration options for the kernel currently > > running: > > On the client, the kernel isn't at all involved in NFS quotas other than > for the usual sendmsg()/recvmsg() syscalls as used by the userspace RPC > library. IOW: all the clever stuff happens in userspace. > > I'd therefore try running 'strace' on the 'quota' command in order to > figure out what is going wrong. I tried that earlier but nothing really jumped out at me - it seemed that the server was being contacted without error. If you're interested you can download a copy of the strace output from a representative client from http://www.physics.adelaide.edu.au/~jwoithe/q-mercury.strace.bz2 I would have attached it but I suspect it might make the message too big for the list. In the meantime I might try comparing the strace from the client (where it doesn't work) and the server (where it does) and see if anything obvious jumps out. Regards jonathan ------------------------------------------------------------------------- This SF.net email is sponsored by DB2 Express Download DB2 Express C - the FREE version of DB2 express and take control of your XML. No limits. Just data. Click to get it now. http://sourceforge.net/powerbar/db2/ _______________________________________________ NFS maillist - NFS@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/nfs