Re: [2.6 patch] PCMCIA mustn't select HAVE_IDE

[Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>]

On Wednesday 16 April 2008, Russell King wrote:
> On Wed, Apr 16, 2008 at 01:10:02AM +0300, Adrian Bunk wrote:
> > On Tue, Apr 15, 2008 at 11:03:45PM +0100, Russell King wrote:
> > > On Wed, Apr 16, 2008 at 12:52:23AM +0300, Adrian Bunk wrote:
> > >...
> > > So this is a only impacting ARM wrt. PCMCIA, and given that ARM supplies
> > > an asm/ide.h, it's _entirely_ reasonable that if a platform has PCMCIA
> > > then it supports IDE.
> > > 
> > > > We could simply always select HAVE_IDE on arm instead of manually 
> > > > setting which platforms could possibly get IDE support (e.g. are there
> > > > any boards with PCI slots for which HAVE_IDE is currently not 
> > > > selected?).
> > > 
> > > You could, if there was a demand for it.  As no one has added that,
> > > I conclude that its less common for people to stick an IDE controller
> > > into a PCI backplane.
> > 
> > People can always enable code for stuff they don't use.
> > 
> > But instead of having 14 ARM platforms plus PCMCIA (which is offered 
> > unconditionally on all ARM platforms...) select HAVE_IDE it's simpler
> > to select it once for all ARM platforms.

Please send me a patch doing this, it should be safe for current IDE tree.

> That would seem logical, but Bart objects to that idea.

I don't remember the background but I think it was needed because of badly
perplexed ide_init_hwif_ports() and friends in <asm/ide.h> vs <linux/ide.h>
(almost all this stuff is gone in IDE tree for 2.6.26)...

Ok, I found the patch:

http://git.kernel.org/?p=linux/kernel/git/tglx/history.git;a=commitdiff;h=4b3b8ee5db374b76608537e061f2efd90e21179d

[ tglx's history tree since it is from May 2004. ]

> However, consider that we're gradually transitioning over to being
> exclusively libata only.
> 
> > > In fact, there are only three classes of ARM platforms which have PCI
> > > selected but not HAVE_IDE - IOP13xx, IXP2000, and Orion.  I suspect
> > > the only reason they don't select it because they now use the ATA code
> > > rather than the old IDE code - that's certainly true of Orion.
> > 
> > The libata options are offered unconditionally on all platforms...
> 
> It wasn't *my* choice to restrict IDE on ARM.  See Bart for that
> decision.

It could be that I did the poor job of explaining things back then
but I also didn't like the fact that I needed to restrict the IDE
choice on ARM - the change in question was _necessary_ to start
converting IDE drivers to become real, independent, modular host
drivers and as a preparation for adding proper warm-plug support.

Thanks,
Bart