Return-Path: Received: from mail-vx0-f174.google.com ([209.85.220.174]:65057 "EHLO mail-vx0-f174.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1750874Ab1FHCFp (ORCPT ); Tue, 7 Jun 2011 22:05:45 -0400 Received: by vxi39 with SMTP id 39so38056vxi.19 for ; Tue, 07 Jun 2011 19:05:45 -0700 (PDT) Message-ID: <4DEED8FC.4000000@panasas.com> Date: Tue, 07 Jun 2011 22:05:48 -0400 From: Benny Halevy To: Jim Rees , Trond Myklebust CC: linux-nfs@vger.kernel.org, peter honeyman Subject: Re: [PATCH 88/88] NFS41: do not update isize if inode needs layoutcommit References: <8d4549ecda23f5ada0c88677c9505cba2830dad5.1307464382.git.rees@umich.edu> In-Reply-To: <8d4549ecda23f5ada0c88677c9505cba2830dad5.1307464382.git.rees@umich.edu> Content-Type: text/plain; charset=ISO-8859-1 Sender: linux-nfs-owner@vger.kernel.org List-ID: MIME-Version: 1.0 Better send generic patches separately. This patch needs to go upstream and to stable 2.6.39 Benny On 2011-06-07 13:36, Jim Rees wrote: > From: Peng Tao > > Layout commit is supposed to set server file size similiar to nfs pages. > We should not update client file size for the same reason. > Otherwise we will lose what we have at hand. > > Signed-off-by: Peng Tao > Signed-off-by: Jim Rees > --- > fs/nfs/inode.c | 3 ++- > 1 files changed, 2 insertions(+), 1 deletions(-) > > diff --git a/fs/nfs/inode.c b/fs/nfs/inode.c > index 144f2a3..3f1eb81 100644 > --- a/fs/nfs/inode.c > +++ b/fs/nfs/inode.c > @@ -1294,7 +1294,8 @@ static int nfs_update_inode(struct inode *inode, struct nfs_fattr *fattr) > if (new_isize != cur_isize) { > /* Do we perhaps have any outstanding writes, or has > * the file grown beyond our last write? */ > - if (nfsi->npages == 0 || new_isize > cur_isize) { > + if ((nfsi->npages == 0 && !test_bit(NFS_INO_LAYOUTCOMMIT, &nfsi->flags)) || > + new_isize > cur_isize) { > i_size_write(inode, new_isize); > invalid |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA; > }