Re: [patch 05/10] s390: add missing validation for dasd discipline specific ioctls

[Christoph Hellwig <hch@lst.de>]

On Wed, Feb 08, 2006 at 05:56:59PM +0100, Horst Hummel wrote:
> Christoph Hellwig <hch@lst.de> wrote on 08.02.2006 15:08:39:
> 
> > On Wed, Feb 08, 2006 at 01:37:09PM +0100, Heiko Carstens wrote:
> > > From: Horst Hummel <horst.hummel@de.ibm.com>
> > > 
> > > Because of missing discipline validition dasd ioctls calls may not
> return
> > > when called on a device handled by a different discipline.
> > > (e.g calling ECKD specific BIODASDGATTR on an FBA device).
> > > This addresses one of the issues Christoph has with the dasd driver.
> > 
> > Nack.  the right way to do this is to have per-discipline ioctl
> methods,
> > even if we can't remove the dasd_ioctl_register interface yet due to
> > other disagreements.  Just resurect those parts of my patch.
> > 
> As I already tried to explain, the DASD ioctls are _not_ discipline
> related. A discipline is more or less a 'access method to a physical 
> DASD'. Sine the ioctls are related to 'functional components', each
> ioctl could be valid in combination with none, one, two or every of
> the currently supported disciplines (ECKD, FBA, DIAG).

Right now there's either ioctls that are valid for one (ECKD) or all
disciplines.   As a special case you've put ioctls that are valid for
all disciplines into odd modules, but that's just another case of lots
of crap in s390 drivers that needs sorting out.