xfs
[Top] [All Lists]

Re: [PATCH 4/4] [PATCH 5/5] xfs: fix xfs_quiesce_data

To: Alex Elder <aelder@xxxxxxx>
Subject: Re: [PATCH 4/4] [PATCH 5/5] xfs: fix xfs_quiesce_data
From: Dave Chinner <david@xxxxxxxxxxxxx>
Date: Mon, 14 Sep 2009 10:37:39 +1000
Cc: Christoph Hellwig <hch@xxxxxxxxxxxxx>, xfs@xxxxxxxxxxx
In-reply-to: <1AB9A794DBDDF54A8A81BE2296F7BDFE83ABF5@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx>
References: <20090827231558.701516609@xxxxxxxxxxxxxxxxxxxxxx> <1AB9A794DBDDF54A8A81BE2296F7BDFE83ABF5@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx>
User-agent: Mutt/1.5.18 (2008-05-17)
On Fri, Aug 28, 2009 at 06:18:45PM -0500, Alex Elder wrote:
> Christoph Hellwig wrote:
> > From: Dave Chinner <david@xxxxxxxxxxxxx>
> > 
> > We need to do a synchronous xfs_sync_fsdata to make sure the superblock
> > actually is on disk when we return.
> > 
> > Also remove SYNC_BDFLUSH flag to xfs_sync_inodes because that particular
> > flag is never checked.
> > 
> > Move xfs_filestream_flush call later [hch: why?  seems unrelated].
> 
> I concur with your question.  Why not release the inode references early?

Because if you drop the filestreams references before you write the
data, the data doesn't get put where the filestreams allocator
decided it should go when it created the reference....

Cheers,

Dave.
-- 
Dave Chinner
david@xxxxxxxxxxxxx

<Prev in Thread] Current Thread [Next in Thread>
  • Re: [PATCH 4/4] [PATCH 5/5] xfs: fix xfs_quiesce_data, Dave Chinner <=