Re: [PATCH #upstream-fixes] ata_piix: save, use saved and restore IOCFG

[Andreas Mohr <andi@lisas.de>]

Hi,

On Fri, Jan 02, 2009 at 05:49:17PM +0900, Tejun Heo wrote:
> Hello,
> 
> Andreas Mohr wrote:
> >> This fixes bz#11879.  Andreas Mohr reported and diagnosed the problem.
> > 
> > I'm mighty unhappy ;-)
> > 
> > First, I still think prime cause was a weak disk implementation of Word 93
> > and not BIOS ACPI handling itself (bug #12202 is a PATA SSD, too!).
> > (unless one thinks that BIOS should know about SSD variants of PATA
> > and actively do special-case them itself)
> 
> Frankly, I don't care one way or the other.  All ->cable_detect() is
> supposed to return is the cable type as detected by the controller and
> _STM is not supposed to alter the state no matter what.  I don't think
> delving into _STM implementation and finding out the exact cause of
> flipping cable detection bit leads to the correct solution.  It might
> be caused by PATA SSD not setting the cable bit this time but the next
> BIOS might as well get it wrong for different reason.  Plus, I don't
> see how IDENTIFY data can affect the cable bit unless the ACPI
> implementation is snooping IDENTIFY replies.

Right, viewn from this angle (of preventing _any_ incorrect messing with
cable type during _STM) it makes sense since it's a more generic
solution.

> > Second, you've been keeping silent about the duplicate processing
> > for too long (I didn't know about it at all until marked duplicate),
> > thus nobody else could derive any hard facts from the doubled information.
> 
> My fault but nothing intentional.  When I got the second report, I
> couldn't really remember your report other than the fact that I had a
> similar report which didn't lead to resolution at the time and between
> Christmas and New Year's day, I wasn't paying much attention to bugs
> other than following up on each one as comments come up.  ie. I didn't
> bother to look up which one was the other one, so the late
> association.

...and there are actually some advantages when _not_ revealing this
(comment #30 at bug #11879 for details).

> > Third, it was not just me who reported it, Carl Michal did >= 10
> > reports in his bug.
> 
> Are all of those SSD too?

That STEC PATA one at least, yes.

Thanks,

Andreas Mohr