From mboxrd@z Thu Jan 1 00:00:00 1970 From: Rusty Lynch Date: Tue Mar 23 23:59:07 2004 Subject: [Ocfs2-devel] [PATCH]Enable 2.6 file data flushing In-Reply-To: <20040324042306.GE1516@ca-server1.us.oracle.com> References: <200403240411.i2O4BrRr014124@penguin.co.intel.com> <20040324042306.GE1516@ca-server1.us.oracle.com> Message-ID: <20040324055857.GA14337@penguin.co.intel.com> List-Id: MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit To: ocfs2-devel@oss.oracle.com On Tue, Mar 23, 2004 at 08:23:06PM -0800, Wim Coekaerts wrote: > hmm now I wonder if we dont' also need that in 2.4... > > > On Tue, Mar 23, 2004 at 08:11:53PM -0800, Rusty Lynch wrote: > > The following small patch makes a 2.6 built kernel flush it's > > data buffers to disk on ocfs_sync_inode, fixing bug #46 > > > > http://oss.oracle.com/bugzilla/show_bug.cgi?id=46 > > > > --rusty > > the equivilant call in 2.4 would be filemap_fdatasync(). I gave it a try and it does fix the mmap issue talked about in bug #50. Here is a patch Index: src/file.c =================================================================== --- src/file.c (revision 807) +++ src/file.c (working copy) @@ -68,8 +68,10 @@ { int status; #if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,0) + filemap_fdatawrite(inode->i_mapping); status = sync_mapping_buffers(inode->i_mapping); #else + filemap_fdatasync(inode->i_mapping); status = fsync_inode_buffers(inode); #if LINUX_VERSION_CODE >= KERNEL_VERSION(2,4,18) if (!status)