On Mon, Jan 14, 2002 at 10:59:02AM -0800, Bill Studenmund wrote:
> On Sun, 13 Jan 2002, Bill Sommerfeld wrote:
> 
> > > Surely there is a 'page modified' bit - and the write should only
> > > happen if it is modified?
> >
> > this may just be updating the vnode 'last access' time.
> >
> > if you mount -o noatime,nodevmtime the disk should be more likely to
> > spin down.
> 
> What would be nice is if we could pass info about the spun up/down state
> further up the kernel, so we could create a class of operations which
> happen when the disk is spun up, but not when it's spun down. That way we
> could update atime when the disk is up, but not when it's down.

Doesn't work for ATA disks - as I understand, the disk handles the stuff
itself, you can only set the timeouts for the downpower transitions, and
it powers up automatically (at least if "down" wasn't too "down").

Regards,
	-is