Re: i_version, NFSv4 change attribute

["J. Bruce Fields" <bfields@fieldses.org>]

On Mon, Nov 23, 2009 at 01:11:19PM -0500, Trond Myklebust wrote:
> On Mon, 2009-11-23 at 11:44 -0500, J. Bruce Fields wrote: 
> > If the side we want to optimize is the modifications, I wonder if we
> > could do all the i_version increments on *read* of i_version?:
> > 
> > 	- writes (and other inode modifications) set an "i_version_dirty"
> > 	  flag.
> > 	- reads of i_version clear the i_version_dirty flag, increment
> > 	  i_version, and return the result.
> > 
> > As long as the reader sees i_version_flag set only after it sees the
> > write that caused it, I think it all works?
> 
> That probably won't make much of a difference to performance. Most NFSv4
> clients will have every WRITE followed by a GETATTR operation in the
> same compound, so your i_version_dirty flag will always immediately get
> cleared.

I was only thinking about non-NFS performance.

> The question is, though, why does the jbd2 machinery need to be engaged
> on _every_ write?

Is it?

I thought I remembered a journaling issue from previous discussions, but
Ted seemed concerned just about the overhead of an additional
spinlock, and looking at the code, the only test of I_VERSION that I can
see indeed is in ext4_mark_iloc_dirty(), and indeed just takes a
spinlock and updates the i_version.

--b.

> The NFS clients don't care if we lose an i_version count due to a
> sudden server reboot, since that will trigger a rewrite of the dirty
> data anyway once the server comes back up again.  As long as the
> i_version is guaranteed to be written to stable storage on a
> successful call to fsync(), then the NFS data integrity requirements
> are fully satisfied.