Re: WD_QUIRK_FORCE_LBA48

[Manuel Bouyer <bouyer%antioche.eu.org@localhost>]

On Wed, Dec 16, 2009 at 12:40:54AM +0100, Christoph Badura wrote:
> On Tue, Dec 15, 2009 at 07:35:42PM +0100, Manuel Bouyer wrote:
> > > Why is that wrong?  *All* sectors (from 0 to max capacity) are accessible 
> > > in
> > > LBA48 mode with LBA48 commands. 
> > Exept if the controller doesn't support LBA48.
> 
> Yeah, and if the drive actually reports 2^28 LBA28 accessible sectors instead
> of 2^28-1 as most versions of the spec requires.

most but not all.

> 
> I've asked you multiple times: which drives and controllers exhibit that
> behaviour.  Please give me links where I can find proof that this actually
> exists and is a problem.

Drives: most drives manufactured before 2004 with more than 138Gb I guess.
The fact that it took 2 years for the problem to show up is enough
of a proof.

Controllers: I suspect some earlier promise controllers, but I have
no way to check.

> 
> > Yes. As I already said we got the sector 0xfffffff problem 2 years after
> > we added LBA48 support. So the earlier LBA48 drives probably used
> > 0x10000000 and not 0xfffffff for the max LBA28 capacity.
> 
> You haven't provided any evidence of that.

The fact that our driver worked with large drives before 2004 is enough.

> The driver violates the
> specification because it tries to access sectors beyond 2^28-1 in LBA28 mode

Not because it tries to access sector beyond 2^28-1, but because
it tries to access sectors beyond (words 60-61)-1.

> and it fails to recover when the drives reports an error about that.  I've
> posted proof of that and suggested fixes.  The behaviour you cite could as
> easily explained by the erroneous behaviour of our buggy driver.  So please
> give me concrete evidence to the contrary. 
> 
> > What this changes is that systems with controller which don't understand
> > LBA48 can still access the last sector of the drive.
> > Otherwise we could just use LBA48 for the whole drive.
> 
> You know.  I'm not going to waste any more time on this.  Please see to it,
> that the above two bugs that I have mentioned and provided corrections for
> are fixed.
> 
> Now, in case there are actually people with non-LBA48 capable IDE controllers
> and LBA48 capable IDE drives that are bothered that really can't access
> the last sector out of 268435456 sectors, I make the following promise:
> if there is actually a single person for whom it is important enough to access
> that last sector that they will ship a drive and IDE controller that exhibit
> the behaviour to me, then I will fix the access to that last sector for them.

And in the meantime, they'll have lost their data.
No, we should use the specs in the correct way, and use
word 60-61 instead of a constant for the LBA48 barrier.
The specs themselve are not consistent on this constant.

-- 
Manuel Bouyer <bouyer%antioche.eu.org@localhost>
     NetBSD: 26 ans d'experience feront toujours la difference
--