"Brian Grayson (home)" wrote:

>   On my router box, I have my partitions mounted synchronously, so
> that hopefully whenever it is turned on, it will auto-fsck itself if
> needed with no problems.  This makes it a no-maintenance router, which
> is good considering my wife's Unix background...  :)
> 
>   My thinking was, if I mount synchronously, then at most one
> transaction was outstanding, and thus fsck should be able to fix
> everything by itself.  However, I just ran a series of around 8
> pull-the-plugs.  A few times there were no problems, a few times there
> was a single file or block count that was bad, but twice there were
> multiple fsck errors (two INCORRECT BLOCK COUNT errors, 3 were UNREF
> files, and then the expected resulting triple of bitmap, free blk counts,
> and summary information bad).  The multiple UNREF files etc. clashes with
> my naive understanding of things...
> 
>   So, am I misunderstanding what synchronous mode means?
> 
>   (I realize that I shouldn't need to mount synchronously, but
> ordinary no-softdep-mounts have gone to manual fsck often enough on
> this box that it makes the box not reliable as a turnkey solution for
> folks who don't appreciate Unix.)

You're disk probably has write-caching enabled.  If I recall correctly,
just about all (modern?) IDE disks by default enable write-caching.

Simon.
--
Simon Burge                            <simonb@wasabisystems.com>
NetBSD CDs, Support and Service:    http://www.wasabisystems.com/