[GIT PULL] MCPM backends update

Olof Johansson olof at lixom.net
Sun Aug 4 16:02:01 EDT 2013

On Mon, Jul 22, 2013 at 04:02:45PM -0400, Nicolas Pitre wrote:
> Hello gents,

Hi, sorry for the delayed reply to this.

> Would you please pull the following into ARM-SOC:
> 	git://git.linaro.org/people/nico/linux mcpm+tc2
> This contains
> - Fixes to the existing Vexpress DCSCB backend.
> - Lorenzo's minimal SPC driver required by the TC2 MCPM backend.
>   As discussed on the list, this is not considered to be a MFD driver.
>   This hardly qualify as a bus driver either.  Conceptually, this is
>   a very platform specific register multiplex driver allowing subsystem
>   drivers to hook into.  I therefore moved it to 
>   drivers/platform/vexpress/ which seems to please everyone.
> - The MCPM backend enabling SMP secondary boot and CPU hotplug
>   on the VExpress TC2 big.LITTLE platform.
> - MCPM suspend method to the TC2 backend allowing basic CPU 
>   idle/suspend.  The cpuidle driver that hooks into this will be 
>   submitted separately.
> This is based on v3.11-rc2.  The diffstat follows:
>  .../devicetree/bindings/arm/vexpress-spc.txt       |  36 +++
>  arch/arm/mach-vexpress/Kconfig                     |   9 +
>  arch/arm/mach-vexpress/Makefile                    |   1 +
>  arch/arm/mach-vexpress/dcscb.c                     |  58 ++--
>  arch/arm/mach-vexpress/tc2_pm.c                    | 304 +++++++++++++++++++++
>  drivers/platform/Kconfig                           |   4 +-
>  drivers/platform/Makefile                          |   1 +
>  drivers/platform/vexpress/Kconfig                  |   9 +
>  drivers/platform/vexpress/Makefile                 |   1 +
>  drivers/platform/vexpress/spc.c                    | 253 +++++++++++++++++
>  include/linux/vexpress.h                           |  17 ++
>  11 files changed, 671 insertions(+), 22 deletions(-)

The vexpress code continues to be uncooperative in fitting into any of
the existing conventions of how drivers and platform code is organized
in the kernel, which makes it hard to find a good home for it.

I sort of see the appeal of having a generic driver for this, but the code is
already TC2-specific.

Also, most drivers under drivers/platform are more or less glue code
that sets up other drivers. I don't think we want to have a calling
dependency from mach-vexpress into drivers/platform like this.

If it wasn't for that tight coupling then I think drivers/platform/arm
could be a good location for it.

I'm sure there'll come a time when a similar driver is needed for
something arm64-based on vexpress. Until then I think we're best off
just moving the code under mach-vexpress.

Nico, would you mind moving it once again and this time to
mach-vexpress? I know you already brought it out of drivers/mfd.



More information about the linux-arm-kernel mailing list