On Jun 10, Manuel Bouyer wrote
> Looks like there's none,  at last in ata-3 (and my drive is definitively not
> and ata-4). Looks like disabling the write cache is the only option.
> I'll look at the ata-4 standarts (at work) tomorow morning.
> However, I don't think this can hurt performances in a significant manner.
> I'll let know the results here.
> 

Ok, I hacked my driver to disable the write cache: performances are
much slower (tested with a 'dd if=/dev/zero of=/dev/rwd0b bs=64k count=500':
14s with write cache, 21s without). But this solves the problem.
I don't think this performance loss is acceptable.
ATA-3 doesn't have a 'flush cache' command. ATA-4 has one, but it is marked
optional. So we can't rely on it.
Erik E. Fair suggested that we disable the write cache at unmount time,
and re-enable it at 'mount -rw'. I'll look at how to implement this.

--
Manuel Bouyer, LIP6, Universite Paris VI.           Manuel.Bouyer@lip6.fr
--