[RFC PATCH] ARM: initial multiplatform support

[robherring2@gmail.com (Rob Herring)]

On 08/24/2012 04:27 PM, Olof Johansson wrote:
> Hi,
> 
> A few comments below.
> 
> On Fri, Aug 24, 2012 at 02:50:52PM -0500, Rob Herring wrote:
>> From: Arnd Bergmann <arnd@arndb.de>
>>
>> This lets us build a multiplatform kernel for experimental purposes.
>> However, it will not be useful for any real work, because it relies
>> on a number of useful things to be disabled for now:
>>
>> * SMP support is turned off because of conflicting symbols.
>>   Marc Zyngier has proposed a solution by adding a new SOC
>>   operations structure to hold indirect function pointers
>>   for these, but that work is currently stalled
> 
> On what? Review? or something else? It'd be nice to get some movement
> going on it again. I guess we'll bring it up next week.

It tends to get sent out again during/after every conference. Marc would
be best to comment on the status.

>> * We turn on SPARSE_IRQ unconditionally, which is not supported
>>   on most platforms. Each of them is currently in a different
>>   state, but most are being worked on.
>>
>> * A common clock framework is in place since v3.4 but not yet
>>   being used. Work on this is on its way.
>>
>> * DEBUG_LL for early debugging is currently disabled.
> 
> Being able to turn it on for one platform just to debug a boot issue is
> still very useful. It's how powerpc handles it -- even if you can do a
> multiplatform kernel, for the low level debug console you have to pick
> one static entry.  Still useful during development.

This comment is misleading. It is possible to select it, but then you
just can only use that 1 platform.

>> * THUMB2_KERNEL does not work with allyesconfig because the
>>   kernel gets too big
> 
> Ironic, but understandable.
> 
>> ---
>>  arch/arm/Kconfig                              |  107 +++++++++++++++----------
>>  arch/arm/Makefile                             |   16 ++--
>>  arch/arm/mach-multi/Kconfig                   |   42 ++++++++++
>>  arch/arm/mach-multi/Makefile                  |    1 +
>>  arch/arm/mach-multi/include/mach/gpio.h       |    1 +
>>  arch/arm/mach-multi/include/mach/timex.h      |    3 +
>>  arch/arm/mach-multi/include/mach/uncompress.h |   10 +++
>>  arch/arm/mach-vexpress/Makefile               |    1 +
>>  arch/arm/plat-versatile/Makefile              |    2 +
>>  9 files changed, 136 insertions(+), 47 deletions(-)
>>  create mode 100644 arch/arm/mach-multi/Kconfig
>>  create mode 100644 arch/arm/mach-multi/Makefile
>>  create mode 100644 arch/arm/mach-multi/include/mach/gpio.h
>>  create mode 100644 arch/arm/mach-multi/include/mach/timex.h
>>  create mode 100644 arch/arm/mach-multi/include/mach/uncompress.h
>>
>> diff --git a/arch/arm/Kconfig b/arch/arm/Kconfig
>> index e91c7cd..e841262 100644
>> --- a/arch/arm/Kconfig
>> +++ b/arch/arm/Kconfig
>> @@ -255,22 +255,18 @@ choice
>>  	prompt "ARM system type"
>>  	default ARCH_VERSATILE
>>  
>> -config ARCH_SOCFPGA
>> -	bool "Altera SOCFPGA family"
>> -	select ARCH_WANT_OPTIONAL_GPIOLIB
>> -	select ARM_AMBA
>> -	select ARM_GIC
>> -	select CACHE_L2X0
>> -	select CLKDEV_LOOKUP
>> +config ARCH_MULTIPLATFORM
>> +	bool "Allow multiple platforms to be selected"
>> +	select USE_OF
> 
> Should this be a toplevel config option and the choice menu disabled if it's
> set, instead of a meta-platform?

If we get rid of mach-multi, yes.

>> @@ -1028,6 +999,58 @@ config ARCH_ZYNQ
>>  	  Support for Xilinx Zynq ARM Cortex A9 Platform
>>  endchoice
>>  
>> +source "arch/arm/mach-multi/Kconfig"
>> +
>> +config ARCH_SOCFPGA
>> +	bool "Altera SOCFPGA family" if ARCH_MULTI_V7
>> +	select ARCH_WANT_OPTIONAL_GPIOLIB
>> +	select ARM_AMBA
>> +	select ARM_GIC
>> +	select CACHE_L2X0
>> +	select CLKDEV_LOOKUP
>> +	select COMMON_CLK
>> +	select CPU_V7
>> +	select DW_APB_TIMER
>> +	select DW_APB_TIMER_OF
>> +	select GENERIC_CLOCKEVENTS
>> +	select GPIO_PL061 if GPIOLIB
>> +	select HAVE_ARM_SCU
>> +	select SPARSE_IRQ
>> +	select USE_OF
> 
> There's no need to keep these in the common arm/Kconfig now, is there? Might as
> well push them out to the respective mach-*/Kconfig, that will cause less churn
> on the common file and in general scale better.

Good point. That's better than creating a Kconfig.mach. This can
probably be done for all platforms independent of multi-platform support.

>> diff --git a/arch/arm/Makefile b/arch/arm/Makefile
>> index 30eae87..c113e53 100644
>> --- a/arch/arm/Makefile
>> +++ b/arch/arm/Makefile
>> @@ -133,6 +133,9 @@ textofs-$(CONFIG_ARCH_MSM7X30) := 0x00208000
>>  textofs-$(CONFIG_ARCH_MSM8X60) := 0x00208000
>>  textofs-$(CONFIG_ARCH_MSM8960) := 0x00208000
>>  
>> +# multiplatform directory must be first
>> +machine-$(CONFIG_ARCH_MULTIPLATFORM)	:= multi
>> +
>>  # Machine directory name.  This list is sorted alphanumerically
>>  # by CONFIG_* macro name.
>>  machine-$(CONFIG_ARCH_AT91)		:= at91
>> @@ -145,7 +148,7 @@ machine-$(CONFIG_ARCH_EBSA110)		:= ebsa110
>>  machine-$(CONFIG_ARCH_EP93XX)		:= ep93xx
>>  machine-$(CONFIG_ARCH_GEMINI)		:= gemini
>>  machine-$(CONFIG_ARCH_H720X)		:= h720x
>> -machine-$(CONFIG_ARCH_HIGHBANK)		:= highbank
>> +machine-$(CONFIG_ARCH_HIGHBANK)		+= highbank
>>  machine-$(CONFIG_ARCH_INTEGRATOR)	:= integrator
>>  machine-$(CONFIG_ARCH_IOP13XX)		:= iop13xx
>>  machine-$(CONFIG_ARCH_IOP32X)		:= iop32x
>> @@ -186,11 +189,11 @@ machine-$(CONFIG_ARCH_TEGRA)		:= tegra
>>  machine-$(CONFIG_ARCH_U300)		:= u300
>>  machine-$(CONFIG_ARCH_U8500)		:= ux500
>>  machine-$(CONFIG_ARCH_VERSATILE)	:= versatile
>> -machine-$(CONFIG_ARCH_VEXPRESS)		:= vexpress
>> +machine-$(CONFIG_ARCH_VEXPRESS)		+= vexpress
>>  machine-$(CONFIG_ARCH_VT8500)		:= vt8500
>>  machine-$(CONFIG_ARCH_W90X900)		:= w90x900
>>  machine-$(CONFIG_FOOTBRIDGE)		:= footbridge
>> -machine-$(CONFIG_ARCH_SOCFPGA)		:= socfpga
>> +machine-$(CONFIG_ARCH_SOCFPGA)		+= socfpga
>>  machine-$(CONFIG_MACH_SPEAR1310)	:= spear13xx
>>  machine-$(CONFIG_MACH_SPEAR1340)	:= spear13xx
>>  machine-$(CONFIG_MACH_SPEAR300)		:= spear3xx
>> @@ -230,14 +233,17 @@ MACHINE  := arch/arm/mach-$(word 1,$(machine-y))/
>>  else
>>  MACHINE  :=
>>  endif
>> +ifeq ($(CONFIG_ARCH_MULTIPLATFORM),y)
>> +MACHINE  :=
>> +endif
>>  
>>  machdirs := $(patsubst %,arch/arm/mach-%/,$(machine-y))
>>  platdirs := $(patsubst %,arch/arm/plat-%/,$(plat-y))
>>  
>>  ifeq ($(KBUILD_SRC),)
>> -KBUILD_CPPFLAGS += $(patsubst %,-I%include,$(machdirs) $(platdirs))
>> +KBUILD_CPPFLAGS += $(patsubst %,-I%include,$(word 1,$(machdirs) $(platdirs)))
>>  else
>> -KBUILD_CPPFLAGS += $(patsubst %,-I$(srctree)/%include,$(machdirs) $(platdirs))
>> +KBUILD_CPPFLAGS += $(patsubst %,-I$(srctree)/%include,$(word 1,$(machdirs) $(platdirs)))
>>  endif
> 
> Since the only purpose of mach-multi is to provide a few dummy header files,
> how about adding those under arch/arm/include/ instead and do away with the
> mach directory? They could then be prepended in front of these so the search
> path would be in the right order.

Yes, good idea.

>> diff --git a/arch/arm/mach-multi/Kconfig b/arch/arm/mach-multi/Kconfig
>> new file mode 100644
>> index 0000000..0a6369f
>> --- /dev/null
>> +++ b/arch/arm/mach-multi/Kconfig
>> @@ -0,0 +1,42 @@
>> +menu "Multiple platform selection"
>> +	depends on ARCH_MULTIPLATFORM
>> +
>> +comment "CPU Core family selection"
>> +
>> +config ARCH_MULTI_V4
>> +	bool "ARMv4 based platforms (FA526, StrongARM)"
>> +	select ARCH_MULTI_V4_V5
>> +	depends on !ARCH_MULTI_V6_V7
>> +
>> +config ARCH_MULTI_V4T
>> +	bool "ARMv4T based platforms (ARM720T, ARM920T, ...)"
>> +	select ARCH_MULTI_V4_V5
>> +	depends on !ARCH_MULTI_V6_V7
>> +
>> +config ARCH_MULTI_V5
>> +	bool "ARMv5 based platforms (ARM926T, XSCALE, PJ1, ...)"
>> +	select ARCH_MULTI_V4_V5
>> +	depends on !ARCH_MULTI_V6_V7
>> +
>> +config ARCH_MULTI_V4_V5
>> +	bool
>> +
>> +config ARCH_MULTI_V6
>> +	bool "ARMv6 based platforms (ARM11, Scorpion, ...)"
>> +	select ARCH_MULTI_V6_V7
> 
> Does it make sense to add these before any platforms are ready? I'd argue no,
> but I don't feel strongly about it.

I'll defer that to Arnd. I'm not sure where he was going with the AUTO
selection stuff.

>> +config ARCH_MULTI_V7
>> +	bool "ARMv7 based platforms (Cortex-A, PJ4, Krait)"
>> +	default y
>> +	select ARCH_MULTI_V6_V7
>> +
>> +config ARCH_MULTI_V6_V7
>> +	bool
>> +
>> +config ARCH_MULTI_CPU_AUTO
>> +	def_bool !(ARCH_MULTI_V4 || ARCH_MULTI_V4T || ARCH_MULTI_V6_V7)
>> +	select ARCH_MULTI_V5
>> +
>> +endmenu
>> +
>> +comment "SoC family selection"
>> diff --git a/arch/arm/mach-multi/Makefile b/arch/arm/mach-multi/Makefile
>> new file mode 100644
>> index 0000000..1bb8bf6
>> --- /dev/null
>> +++ b/arch/arm/mach-multi/Makefile
>> @@ -0,0 +1 @@
>> +# empty
>> diff --git a/arch/arm/mach-multi/include/mach/gpio.h b/arch/arm/mach-multi/include/mach/gpio.h
>> new file mode 100644
>> index 0000000..40a8c17
>> --- /dev/null
>> +++ b/arch/arm/mach-multi/include/mach/gpio.h
>> @@ -0,0 +1 @@
>> +/* empty */
>> diff --git a/arch/arm/mach-multi/include/mach/timex.h b/arch/arm/mach-multi/include/mach/timex.h
>> new file mode 100644
>> index 0000000..929b4c4
>> --- /dev/null
>> +++ b/arch/arm/mach-multi/include/mach/timex.h
>> @@ -0,0 +1,3 @@
>> +#ifndef CLOCK_TICK_RATE
>> +#define CLOCK_TICK_RATE 1000000
>> +#endif
>> diff --git a/arch/arm/mach-multi/include/mach/uncompress.h b/arch/arm/mach-multi/include/mach/uncompress.h
>> new file mode 100644
>> index 0000000..a4dec68
>> --- /dev/null
>> +++ b/arch/arm/mach-multi/include/mach/uncompress.h
>> @@ -0,0 +1,10 @@
>> +#ifndef __MACH_UNCOMPRESS_H
>> +#define __MACH_UNCOMPRESS_H
>> +
>> +/* no-op dummies for the uncompress debugging for multiplatform kernels */
>> +
>> +static inline void putc(int c) {}
>> +static inline void flush(void) {}
>> +static inline void arch_decomp_setup(void) {}
>> +
>> +#endif
>> diff --git a/arch/arm/mach-vexpress/Makefile b/arch/arm/mach-vexpress/Makefile
>> index 90551b9..1666f6e 100644
>> --- a/arch/arm/mach-vexpress/Makefile
>> +++ b/arch/arm/mach-vexpress/Makefile
>> @@ -1,6 +1,7 @@
>>  #
>>  # Makefile for the linux kernel.
>>  #
>> +ccflags-y := -I$(srctree)/$(src)/include -I$(srctree)/arch/arm/plat-versatile/include
> 
> cc-flags-$(CONFIG_ARCH_MULTIPLATFORM) here and for versatile?

Yes. I'd really like to find a way to do this in arch/arm/Makefile instead.

Rob