[PATCH] ARM: Dove: allow PCI to be disabled

[Jason Cooper]

On Mon, Sep 10, 2012 at 01:45:11AM +0100, Russell King - ARM Linux wrote:
> On Sun, Sep 09, 2012 at 05:01:26PM -0400, Jason Cooper wrote:
> > On Sun, Sep 09, 2012 at 09:55:21PM +0100, Russell King - ARM Linux wrote:
> > > On Sun, Sep 09, 2012 at 10:47:05PM +0200, Andrew Lunn wrote:
> > > > On Sun, Sep 09, 2012 at 08:34:13PM +0100, Russell King wrote:
> > > > > Allow PCI support for Dove to be disabled.  Some platforms do not have
> > > > > anything connected to the PCIe ports, so requiring PCI support to be
> > > > > built into the kernel just wastes space, and presents a lot more config
> > > > > options than are necessary.
> > > > > 
> > > > > However, select USB_ARCH_HAS_EHCI so that we can still have EHCI
> > > > > support.
> > > > 
> > > > Hi Russel
> > > > 
> > > > I have a patch in GregKH usb-next which also ensures USB_ARCH_HAS_EHCI
> > > > is selected.
> > > > 
> > > > http://git.kernel.org/?p=linux/kernel/git/gregkh/usb.git;a=commitdiff;h=ba2f9dff6c914b91005f687a9d75c8eac110d323
> > > > 
> > > > So we might not need your select as well, although it does not harm.
> > > 
> > > That depends on how this patch is handled.  If it gets into mainline via
> > > a separate path to GregKH's USB tree, then it must be kept together as
> > > otherwise it _will_ cause regressions - namely EHCI becomes impossible
> > > to select when PCI=n.  That in turn makes the CuBox (and probably other
> > > Dove platforms) a rather useless di
> > [message trucated, I need to working on file locking for my email
> > auto-flagging script... :-/]
> > 
> > greg just pulled Andrew's patch into his usb tree within the past few
> > days.  Russell, would you prefer to submit this to greg's usb tree, or
> > resend without USB_ARCH_HAS_EHCI?
> 
> I thought I'd already answered the latter point.
> 
> As I said, it breaks the patch to remove that select, so it can not be
> removed.  Removing it will create regressions - can I be any more clear
> about that.  So no, I will not remove that select statement because I
> can't without _breaking_ my Dove builds here.

My apologies.  I was simply trying to achieve the cleanest solution.
Andrew's patch selects USB_ARCH_HAS_EHCI whenever PLAT_ORION is enabled,
of which dove is one.  Purely from the pov of the code, this makes the
select redundant.

iiuc, the regression you are refering to would occur iff this patch
(without the select) makes it into mainline _before_ Andrew's patch in
Greg's tree.  As I haven't pulled this patch yet, I can add a dependency
in the tag message listing the patch in gregkh's usb tree.  Arnd and
Olof would heed this and make sure this patch comes after Andrews.  This
should prevent the regression.

If the above is true, then indeed, the select is not necessary.  Is
there something else I'm missing?

> I don't think submitting this patch to Greg's USB tree is also
> appropriate either, it's clearly ARM specific in nature, or even PCI
> specific.  It isn't a USB patch.

Agreed, please chalk that up to new maintainer crack-smoking. :-)

> Neither is a patch changing a USB setting in arch/arm/Kconfig.  If any
> tree, these patches should be going through arm-soc.

Agreed.

thx,

Jason.