Chris Wright wrote:
> * Zachary Amsden (zach@vmware.com) wrote:
>   
>> Comments, suggestions, anything welcome.  I think this is a much cleaner 
>> approach, and both new and existing sub-architectures will benefit.  I 
>> am sorry this patch is so large, but it is very difficult to separate 
>> into multiple steps that still allow all the subarches to compile.
>>     

Thanks for looking over these.  There is a bit more cleanup on top of 
this, mostly due to the include mess.  I have attached another patch.

> Zach, looks nice.  Saves Xen a partial copy of setup.c.  Did you have
> further/similar consolidations in mind?
>   

Exactly.
>   
>> --- linux-2.6.16.1.orig/arch/i386/Makefile	2006-03-29 19:38:47.000000000 -0800
>> +++ linux-2.6.16.1/arch/i386/Makefile	2006-03-29 19:38:54.000000000 -0800
>> @@ -45,37 +45,32 @@ CFLAGS				+= $(shell if [ $(call cc-vers
>>  
>>  CFLAGS += $(cflags-y)
>>  
>> -# Default subarch .c files
>> -mcore-y  := mach-default
>> +# Default subarch .c files (none)
>> +mcore-y  := 
>>  
>>  # Voyager subarch support
>>  mflags-$(CONFIG_X86_VOYAGER)	:= -Iinclude/asm-i386/mach-voyager
>> -mcore-$(CONFIG_X86_VOYAGER)	:= mach-voyager
>> +mcore-$(CONFIG_X86_VOYAGER)	:= arch/i386/mach-voyager/
>>     
>
> Is this intended to make way for possible fine tuning?  Smth like:
>
> mcore-$(CONFIG_X86_VOYAGER)	+= arch/i386/another_default.o
> (hmm, not sure if that would even work)
>
> Or just an aesthetic change?
>   

No - it was required in order to leave the subarch blank for the default 
case.  Including "arch/i386//" wasn't so happy in the makefile.

>   
>> --- linux-2.6.16.1.orig/include/asm-i386/acpi.h	2006-03-29 19:38:47.000000000 -0800
>> +++ linux-2.6.16.1/include/asm-i386/acpi.h	2006-03-29 19:38:54.000000000 -0800
>> @@ -31,6 +31,7 @@
>>  #include <acpi/pdc_intel.h>
>>  
>>  #include <asm/system.h>		/* defines cmpxchg */
>> +#include <asm/processor.h>	/* defines boot_cpu_data */
>>     
>
> that one necessary?
>   

Needed for mach-generic subarch to compile with CPU defined as I386 -- 
the definition of cmpxchg on older CPUs depends on boot_cpu_data.  
Removing this opens a giant rat hole.  This ginourmous rat hole is quite 
annoying.  What goes in system.h vs. processor.h is so poorly defined.  
I had to hoist boot_cpu_data into system.h as the cleanest way to fix 
this without introducing circular dependencies - processor.h already 
depends on system.h for alternative_input(), read_cr4(), write_cr4(), 
write_cr3(), and there is no end to this mess.  Trying to move dependent 
processor.h pieces into system.h didn't work, since the include/linux 
headers make the assumption that they can include asm/processor.h to 
determine if the architecture already has a prefetch instruction 
defined.  What a nightmare.  Trying to move system.h pieces into 
processor.h also doesn't work, since you basically need to move the 
whole file.

Maybe that is the way to go, but I would rather not go there now.


>>  #define COMPILER_DEPENDENT_INT64   long long
>>  #define COMPILER_DEPENDENT_UINT64  unsigned long long
>> Index: linux-2.6.16.1/include/asm-i386/arch_hooks.h
>> ===================================================================
>> --- linux-2.6.16.1.orig/include/asm-i386/arch_hooks.h	2006-03-29 19:38:47.000000000 -0800
>> +++ linux-2.6.16.1/include/asm-i386/arch_hooks.h	2006-03-29 19:38:54.000000000 -0800
>> @@ -1,7 +1,13 @@
>>  #ifndef _ASM_ARCH_HOOKS_H
>>  #define _ASM_ARCH_HOOKS_H
>>  
>> +#include <linux/config.h>
>> +#include <linux/smp.h>
>> +#include <linux/init.h>
>>  #include <linux/interrupt.h>
>> +#include <asm/acpi.h>
>> +#include <asm/arch_hooks.h>
>>     
>
> extraneous include
>   

Yes, somewhat extraneous.

>   
>> --- linux-2.6.16.1.orig/include/asm-i386/mach-default/mach_hooks.h	2006-03-29 19:38:54.000000000 -0800
>> +++ linux-2.6.16.1/include/asm-i386/mach-default/mach_hooks.h	2006-03-29 19:38:54.000000000 -0800
>> @@ -0,0 +1,6 @@
>> +#ifndef _MACH_HOOKS_H
>> +#define _MACH_HOOKS_H
>>     
>
> should probably be consistent (_MACH_HOOKS_H vs. MACH_HOOKS_H)
>   
Patch attached.