xfs
[Top] [All Lists]

Re: [PATCH 2/2] xfs: log all dirty inodes in xfs_fs_sync_fs

To: Dave Chinner <david@xxxxxxxxxxxxx>
Subject: Re: [PATCH 2/2] xfs: log all dirty inodes in xfs_fs_sync_fs
From: Christoph Hellwig <hch@xxxxxxxxxxxxx>
Date: Sun, 18 Dec 2011 17:32:39 -0500
Cc: Christoph Hellwig <hch@xxxxxxxxxxxxx>, xfs@xxxxxxxxxxx, Paul Anderson <pha@xxxxxxxxx>, Sean Thomas Caron <scaron@xxxxxxxxx>
In-reply-to: <20111218221707.GH23662@dastard>
References: <20111218154936.GA17626@xxxxxxxxxxxxx> <20111218155015.GC17626@xxxxxxxxxxxxx> <20111218221707.GH23662@dastard>
User-agent: Mutt/1.5.21 (2010-09-15)
On Mon, Dec 19, 2011 at 09:17:07AM +1100, Dave Chinner wrote:
> This will do a transaction on the inode, clean or dirty. That's an
> awful lot of overhead for the few inodes (out of perhaps millions in
> memory) that actually need it. with the ->dirty_inode callback from
> the VFS, we know the only inodes that need logging are those with
> i_update_core set....

Ooops, I messed that up when forwarding the RFC patch I sent to
Paul & Sean, and that I had been testing with most of the time.  The
original one moved the i_update_core check into xfs_log_inode and that
is how it should be done.  I'll resend it.

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