[PATCH 10/37] ARM: mach-ep93xx: remove mach/memory.h and Kconfig selection of SDRAM bank

[Nicolas Pitre]

On Thu, 7 Jul 2011, H Hartley Sweeten wrote:

> On Thursday, July 07, 2011 10:19 AM, Nicolas Pitre wrote:
> > On Thu, 7 Jul 2011, H Hartley Sweeten wrote:
> >
> >> On Tuesday, July 05, 2011 8:30 PM, Nicolas Pitre wrote:
> >>>
> >>> This all can be selected automatically by default at run time.
> >> 
> >> How?  It would be nice if you mentioned that this patch makes ARCH_EP93XX
> >> use ARM_PATCH_PHYS_VIRT.  It took a bit of digging for me to figure out
> >> where PHYS_OFFSET was now getting set.
> >
> > OK, I'll add some precisions to the commit message.  Of course the first 
> > patch of the series contains detailed information on what is happening 
> > but I don't want to repeat that into the 51 subsequent patches.
> 
> Ah.. I read the first patch to quickly...  Still an updated commit would
> keep it clear.  Something as simple as this.
> 
> This all can be selected automatically by default at run time using
> ARM_PATCH_PHYS_VIRT.

Yes, I've done something similar.

> >>> Signed-off-by: Nicolas Pitre <nicolas.pitre at linaro.org>
> >>> ---
> >>>  arch/arm/Kconfig                           |    2 +
> >>>  arch/arm/mach-ep93xx/Kconfig               |   53 ----------------------------
> >>>  arch/arm/mach-ep93xx/Makefile.boot         |   14 -------
> >>>  arch/arm/mach-ep93xx/include/mach/memory.h |   22 -----------
> >>>  4 files changed, 2 insertions(+), 89 deletions(-)
> >>>  delete mode 100644 arch/arm/mach-ep93xx/include/mach/memory.h
> >>>
> >>> diff --git a/arch/arm/Kconfig b/arch/arm/Kconfig
> >>> index c4b6e4a..48b2e88 100644
> >>> --- a/arch/arm/Kconfig
> >>> +++ b/arch/arm/Kconfig
> >>> @@ -371,6 +371,8 @@ config ARCH_EP93XX
> >>>  	select ARCH_REQUIRE_GPIOLIB
> >>>  	select ARCH_HAS_HOLES_MEMORYMODEL
> >>>  	select ARCH_USES_GETTIMEOFFSET
> >>> +	select NO_MACH_MEMORY_H
> >>> +	select AUTO_ZRELADDR if !ZBOOT_ROM
> >> 
> >> I would like to see a:
> >> 
> >> +	select ARM_PATCH_PHYS_VIRT
> >> 
> >> here.  I realize that the select of NO_MACH_MEMORY_H will cause the selection
> >> but it's a bit hidden.
> >
> > True.  However the default for ARM_PATCH_PHYS_VIRT is set to y when 
> > NO_MACH_MEMORY_H is selected which effectively achieve the same thing 
> > while allowing you to override that and 
> > revert to a constant PHYS_OFFSET if some regression is suspected, or if 
> > you use a XIP kernel.  Doing a forceful select of ARM_PATCH_PHYS_VIRT 
> > here would prevent that..
> 
> Ah.. I overlooked that part.  Leave it as-is then.
> 
> >> [snip]
> >> 
> >>> diff --git a/arch/arm/mach-ep93xx/Makefile.boot b/arch/arm/mach-ep93xx/Makefile.boot
> >>> index 0ad33f1..e69de29 100644
> >>> --- a/arch/arm/mach-ep93xx/Makefile.boot
> >>> +++ b/arch/arm/mach-ep93xx/Makefile.boot
> >>> @@ -1,14 +0,0 @@
> >>> -   zreladdr-$(CONFIG_EP93XX_SDCE3_SYNC_PHYS_OFFSET)	:= 0x00008000
> >>> -params_phys-$(CONFIG_EP93XX_SDCE3_SYNC_PHYS_OFFSET)	:= 0x00000100
> >>> -
> >>> -   zreladdr-$(CONFIG_EP93XX_SDCE0_PHYS_OFFSET)		:= 0xc0008000
> >>> -params_phys-$(CONFIG_EP93XX_SDCE0_PHYS_OFFSET)		:= 0xc0000100
> >>> -
> >>> -   zreladdr-$(CONFIG_EP93XX_SDCE1_PHYS_OFFSET)		:= 0xd0008000
> >>> -params_phys-$(CONFIG_EP93XX_SDCE1_PHYS_OFFSET)		:= 0xd0000100
> >>> -
> >>> -   zreladdr-$(CONFIG_EP93XX_SDCE2_PHYS_OFFSET)		:= 0xe0008000
> >>> -params_phys-$(CONFIG_EP93XX_SDCE2_PHYS_OFFSET)		:= 0xe0000100
> >>> -
> >>> -   zreladdr-$(CONFIG_EP93XX_SDCE3_ASYNC_PHYS_OFFSET)	:= 0xf0008000
> >>> -params_phys-$(CONFIG_EP93XX_SDCE3_ASYNC_PHYS_OFFSET)	:= 0xf0000100
> >> 
> >> Is Makefile.boot still needed for anything?
> >
> > Not if AUTO_ZRELADDR is used and you are not using the bootpImage 
> > wrapper.
> 
> Ok.  Just seems weird to have an empty Makefile.boot...

Well, the build fails without it.  I've left it empty for now.  A 
subsequent cleanup will eventually remove it.

> > This should allow you to have a kernel configured for multiple ep93xx 
> > boards where this was not possible before due to the different RAM 
> > offsets hardcoded at build time.
> 
> Actually it was "kind" of possible before.  The limitation was that all the
> ep93xx boards needed to use the same phys offset.

Well, before this patch, board choices were dependent on either 
EP93XX_SDCE0_PHYS_OFFSET, EP93XX_SDCE3_SYNC_PHYS_OFFSET, 
EP93XX_SDCE3_ASYNC_PHYS_OFFSET, etc.  Now you should be able to support 
them all together with a single kernel binary now that this distinction 
is gone.

> >> Acked-by: H Hartley Sweeten <hsweeten at visionengravers.com>
> >
> > May I add a Tested-by tag as well?
> 
> Sure...
> 
> Tested-by: H Hartley Sweeten <hsweeten at visionengravers.com>

Thanks.


Nicolas