Re: [RFC PATCH 1/3] arch, arm32: add definition of paddr_bits

[Arianna Avanzini <avanzini.arianna@gmail.com>]

On 03/02/2014 09:13 AM, Julien Grall wrote:
> Hello Arianna,
> 

Hello,

> On 02/03/14 08:49, Arianna Avanzini wrote:
>> Currently, the paddr_bits variable, which keeps the machine word size,
>> is not defined for the arm32 architecture. This commits adds its
>> definition for arm32 in a new header in the architecture-specific
>> directory.
>> This change is instrumental to the implementation of the
>> XEN_DOMCTL_memory_mapping hypercall, added in the following
>> commit.
> 
> paddr_bits doesn't keep the machine word size but the size of the physical address.
> 
> On ARM, this value is fixed and defined as PADDR_BITS in
> xen/include/asm-arm/page.h. I don't think we need to introduce a variable for
> this purpose. x86 defines paddr_bits because the value can change on some
> processor.
> 

Thank you for explaining that. I've tried to fix as suggested.

>> diff --git a/xen/arch/arm/arm32/cpu.h b/xen/arch/arm/arm32/cpu.h
>> new file mode 100644
>> index 0000000..1091051
>> --- /dev/null
>> +++ b/xen/arch/arm/arm32/cpu.h
>> @@ -0,0 +1,7 @@
>> +#ifndef __ARM_ARM32_CPU_H
>> +#define __ARM_ARM32_CPU_H
>> +
>> +/* ARM-v7 features LPAE (machine word size is 40 bit) */
>> +unsigned int paddr_bits = 40;
>> +
> 
> Global variable must not be defined in headers. If someone decides to include
> this header in 2 differents C files, the variable will be defined twice.
> Therefore the linker-step will fail.
> 
> You should define it in a c-file.
> 

OK, thank you again.


> Regards,
> 


-- 
/*
 * Arianna Avanzini
 * avanzini.arianna@gmail.com
 * 73628@studenti.unimore.it
 */