[Top] [All Lists]

Re: [PATCH 11/17] xfs: merge xfs_inode_flush into xfs_fs_write_inode

To: Christoph Hellwig <hch@xxxxxxxxxxxxx>
Subject: Re: [PATCH 11/17] xfs: merge xfs_inode_flush into xfs_fs_write_inode
From: Dave Chinner <david@xxxxxxxxxxxxx>
Date: Mon, 16 Feb 2009 14:06:57 +1100
Cc: xfs@xxxxxxxxxxx
In-reply-to: <20090215201321.GA16048@xxxxxxxxxxxxx>
Mail-followup-to: Christoph Hellwig <hch@xxxxxxxxxxxxx>, xfs@xxxxxxxxxxx
References: <20090126073136.384490000@xxxxxxxxxxxxxxxxxxxxxx> <20090126073202.588218000@xxxxxxxxxxxxxxxxxxxxxx> <20090215064225.GA8830@disturbed> <20090215201321.GA16048@xxxxxxxxxxxxx>
User-agent: Mutt/1.5.18 (2008-05-17)
On Sun, Feb 15, 2009 at 03:13:21PM -0500, Christoph Hellwig wrote:
> On Sun, Feb 15, 2009 at 05:42:25PM +1100, Dave Chinner wrote:
> > On Mon, Jan 26, 2009 at 02:31:47AM -0500, Christoph Hellwig wrote:
> > > Spliting the task for a VFS-induced inode flush into two functions doesn't
> > > make any sense, so merge the two functions dealing with it.
> > 
> > I just realised that we should really be calling xfs_inode_flush() from
> > xfs_sync_inodes_ag() - it open codes xfs_inode_flush() and doesn't
> > have any of the non-blocking flush code for async flushes...
> Given that we'll hopefully do batched flushes driven directly from
> the sync code re-introducing the helper now just to rip it out again
> a little later doesn't seem to useful to me.  And the calling
> conventions for xfs_inode_flush were rather odd to start with anyway.

Well, ok. I'll just duplicate xfs_fs_write_inode() for the sync code
right now, then.


Dave Chinner

<Prev in Thread] Current Thread [Next in Thread>