Johannes Weiner wrote:
> On Thu, Jun 18, 2009 at 01:24:10PM -0700, Andrew Morton wrote:
>> On Thu, 18 Jun 2009 13:07:13 -0700
>> Joe Perches <joe@perches.com> wrote:
>>
>>> Signed-off-by: Joe Perches <joe@perches.com>
>>>
>>> diff --git a/mm/bootmem.c b/mm/bootmem.c
>>> index 282df0a..09d9c98 100644
>>> --- a/mm/bootmem.c
>>> +++ b/mm/bootmem.c
>>> @@ -536,11 +536,13 @@ static void * __init alloc_arch_preferred_bootmem(bootmem_data_t *bdata,
>>>  		return kzalloc(size, GFP_NOWAIT);
>>>  
>>>  #ifdef CONFIG_HAVE_ARCH_BOOTMEM
>>> +	{
>>>  	bootmem_data_t *p_bdata;
>>>  
>>>  	p_bdata = bootmem_arch_preferred_node(bdata, size, align, goal, limit);
>>>  	if (p_bdata)
>>>  		return alloc_bootmem_core(p_bdata, size, align, goal, limit);
>>> +	}
>>>  #endif
>>>  	return NULL;
>>>  }
>> Well yes.
>>
>> We'll be needing some tabs there.
>>
>> Unrelatedly, I'm struggling a bit with bootmem_arch_preferred_node(). 
>> It's only defined if CONFIG_X86_32=y && CONFIG_NEED_MULTIPLE_NODES=y,
>> but it gets called if CONFIG_HAVE_ARCH_BOOTMEM=y.
>>
>> Is this correct, logical and as simple as we can make it??
> 
> x86_32 numa is the only setter of HAVE_ARCH_BOOTMEM.  I don't know why
> this arch has a strict preference/requirement(?) for bootmem on node
> 0.
> 
> I found this mail from Yinghai
> 
>   http://marc.info/?l=linux-kernel&m=123614990906256&w=2
> 
> where he says that it expects all bootmem on node zero but with the
> current code and alloc_arch_preferred_bootmem() failing, we could fall
> back to another node.  Won't this break?  Yinghai?

not sure it is the same problem. the fix was in mainline already.


> 
> Otherwise, could we perhaps use something as simple as this?
> 
> diff --git a/arch/x86/include/asm/mmzone_32.h b/arch/x86/include/asm/mmzone_32.h
> index ede6998..b68a672 100644
> --- a/arch/x86/include/asm/mmzone_32.h
> +++ b/arch/x86/include/asm/mmzone_32.h
> @@ -92,8 +92,7 @@ static inline int pfn_valid(int pfn)
>  
>  #ifdef CONFIG_NEED_MULTIPLE_NODES
>  /* always use node 0 for bootmem on this numa platform */
> -#define bootmem_arch_preferred_node(__bdata, size, align, goal, limit)	\
> -	(NODE_DATA(0)->bdata)
> +#define bootmem_arch_preferred_node (NODE(0)->bdata)
>  #endif /* CONFIG_NEED_MULTIPLE_NODES */
>  
>  #endif /* _ASM_X86_MMZONE_32_H */
> diff --git a/mm/bootmem.c b/mm/bootmem.c
> index 282df0a..0097fa2 100644
> --- a/mm/bootmem.c
> +++ b/mm/bootmem.c
> @@ -528,23 +528,6 @@ find_block:
>  	return NULL;
>  }
>  
> -static void * __init alloc_arch_preferred_bootmem(bootmem_data_t *bdata,
> -					unsigned long size, unsigned long align,
> -					unsigned long goal, unsigned long limit)
> -{
> -	if (WARN_ON_ONCE(slab_is_available()))
> -		return kzalloc(size, GFP_NOWAIT);
> -
> -#ifdef CONFIG_HAVE_ARCH_BOOTMEM
> -	bootmem_data_t *p_bdata;
> -
> -	p_bdata = bootmem_arch_preferred_node(bdata, size, align, goal, limit);
> -	if (p_bdata)
> -		return alloc_bootmem_core(p_bdata, size, align, goal, limit);
> -#endif
> -	return NULL;
> -}
> -
>  static void * __init ___alloc_bootmem_nopanic(unsigned long size,
>  					unsigned long align,
>  					unsigned long goal,
> @@ -553,11 +536,15 @@ static void * __init ___alloc_bootmem_nopanic(unsigned long size,
>  	bootmem_data_t *bdata;
>  	void *region;
>  
> +	if (WARN_ON_ONCE(slab_is_available()))
> +		return kzalloc(size, GFP_NOWAIT);
>  restart:
> -	region = alloc_arch_preferred_bootmem(NULL, size, align, goal, limit);
> +#ifdef bootmem_arch_preferred_node
> +	region = alloc_bootmem_core(bootmem_arch_preferred_node,
> +				size, align, goal, limit);
>  	if (region)
>  		return region;
> -
> +#endif
>  	list_for_each_entry(bdata, &bdata_list, list) {
>  		if (goal && bdata->node_low_pfn <= PFN_DOWN(goal))
>  			continue;
> @@ -636,13 +623,11 @@ static void * __init ___alloc_bootmem_node(bootmem_data_t *bdata,
>  {
>  	void *ptr;
>  
> -	ptr = alloc_arch_preferred_bootmem(bdata, size, align, goal, limit);
> -	if (ptr)
> -		return ptr;
> -
> +#ifndef bootmem_arch_preferred_node
>  	ptr = alloc_bootmem_core(bdata, size, align, goal, limit);
>  	if (ptr)
>  		return ptr;
> +#endif
>  
>  	return ___alloc_bootmem(size, align, goal, limit);
>  }


any reason to kill alloc_arch_preferred_bootmem?

YH