On Sun, Feb 16, 2003 at 02:33:05AM -0600, David Young wrote:
> On Sat, Feb 15, 2003 at 11:25:59PM -0500, Perry E. Metzger wrote:
> > 
> > I find myself wanting to do this for a different reason -- I want to
> > assure on system sleep (think APM or ACPI suspends) that all file
> > systems are synced cleanly.
> 
> Interesting problem.
> 
> Say you do a VFS_SYNC. Is there a way to distinguish blocks written to
> accomplish the VFS_SYNC from the blocks written afterward? I mean, is
> there any notion of a block's "generation," where VFS_SYNC's mark the
> beginning of a new generation? Having that info (I realize it might be
> a tall order for that info!), this is not too complicated, right?

Ding!  Read John Kohl's paper on online backup of LFS using a special
cleaner (also note that this is how WAFL does snapshots).  LFS makes
things like this very, very simple, elegant, and clean.

Thor