Re: [PATCH] Make CONFIG_MIGRATION available for s390

[Christoph Lameter]

Gerald Schaefer wrote:

> Index: linux-2.6/include/linux/migrate.h
> ===================================================================
> --- linux-2.6.orig/include/linux/migrate.h
> +++ linux-2.6/include/linux/migrate.h
> @@ -13,6 +13,7 @@ static inline int vma_migratable(struct 
>  {
>  	if (vma->vm_flags & (VM_IO|VM_HUGETLB|VM_PFNMAP|VM_RESERVED))
>  		return 0;
> +#ifdef CONFIG_NUMA
>  	/*
>  	 * Migration allocates pages in the highest zone. If we cannot
>  	 * do so then migration (at least from node to node) is not
> @@ -22,6 +23,7 @@ static inline int vma_migratable(struct 
>  		gfp_zone(mapping_gfp_mask(vma->vm_file->f_mapping))
>  								< policy_zone)
>  			return 0;
> +#endif
>  	return 1;
>  }

This will extend the number of pages that are migratable and lead to strange
semantics in the NUMA case. There suddenly vma_is migratable will forbid hotplug
to migrate certain pages. 

I think we need two functions:

vma_migratable()	General migratability

vma_policy_migratable()	Migratable under NUMA policies.


> Index: linux-2.6/include/linux/mm.h
> ===================================================================
> --- linux-2.6.orig/include/linux/mm.h
> +++ linux-2.6/include/linux/mm.h
> @@ -193,6 +193,8 @@ struct vm_operations_struct {
>  	 */
>  	struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
>  					unsigned long addr);
> +#endif
> +#ifdef CONFIG_MIGRATION
>  	int (*migrate)(struct vm_area_struct *vma, const nodemask_t *from,
>  		const nodemask_t *to, unsigned long flags);
>  #endif

That wont work since the migrate function takes a nodemask! The point of
the function is to move memory from node to node which is something that you
*cannot* do in a non NUMA configuration. So leave this chunk out.


> Index: linux-2.6/mm/Kconfig
> ===================================================================
> --- linux-2.6.orig/mm/Kconfig
> +++ linux-2.6/mm/Kconfig
> @@ -174,7 +174,7 @@ config SPLIT_PTLOCK_CPUS
>  config MIGRATION
>  	bool "Page migration"
>  	def_bool y
> -	depends on NUMA
> +	depends on NUMA || ARCH_ENABLE_MEMORY_HOTREMOVE
>  	help
>  	  Allows the migration of the physical location of pages of processes
>  	  while the virtual addresses are not changed. This is useful for

Hmmm... Okay. I tried to make MIGRATION as independent of CONFIG_NUMA as possible so hopefully this will work.



--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/