Re: [Patch 2/4] x86 module: merge the rest functions with macros

[Amerigo Wang <amwang@redhat.com>]

Christoph Hellwig wrote:
> On Tue, May 26, 2009 at 04:35:22AM -0400, Amerigo Wang wrote:
>   
>> +#if defined(CONFIG_UML) || defined(CONFIG_X86_32)
>> +void *module_alloc(unsigned long size)
>> +{
>> +	if (size == 0)
>> +		return NULL;
>> +	return vmalloc_exec(size);
>> +}
>> +#else /*X86_64*/
>> +void *module_alloc(unsigned long size)
>> +{
>> +	struct vm_struct *area;
>> +
>> +	if (!size)
>> +		return NULL;
>> +	size = PAGE_ALIGN(size);
>> +	if (size > MODULES_LEN)
>> +		return NULL;
>> +
>> +	area = __get_vm_area(size, VM_ALLOC, MODULES_VADDR, MODULES_END);
>> +	if (!area)
>> +		return NULL;
>> +
>> +	return __vmalloc_area(area, GFP_KERNEL, PAGE_KERNEL_EXEC);
>> +}
>> +#endif
>>     
>
> vmalloc_exec basically expands to the x86-64 version of that code, just
> using VMALLOC_START/VMALLOC_END instead of MODULES_VADDR/MODULES_END.
>
> So instead of having two variants it would be better to use the x86-64
> unconditionally and define MODULES_VADDR/MODULES_END to
> VMALLOC_START/VMALLOC_END to 32bit and uml.
>
> And that part should be a patch of it's own, not mixed with others.
>   

Thanks, it is a good idea!

But... vmalloc_exec() also sets __GFP_HIGHMEM, this is different from 
x86_64.
No? Or __GFP_HIGHMEM is meaningless on x86_64? :-)