Re: [PATCH] mm/Kconfig: allow user to select MIGRATION if MEMORY_FAILURE is enabled

On 8/13/26 04:26, Xie Yuanbin wrote:
> On Wed, 12 Aug 2026 13:45:30 +0300, Mike Rapoport wrote:
>> Do we want to make users decide if they need MIGRATION when MEMORY_FAILURE
>> is enabled?
>>
>> Just make MEMORY_FAILURE select MIGRATION.
> 
> I agree with this solution, but I discovered another way: just like
> commit 6ebf98d71f9b509e833e ("mm: introduce CONFIG_NUMA_MIGRATION and
> simplify CONFIG_MIGRATION") done, add CONFIG_MEMORY_FAILURE_MIGRATION
> to select MIGRATION. The Modification is also very simple:
> ```patch
> diff --git a/mm/memory-failure.c b/mm/memory-failure.c
> index a8b03e2920ba..500e54ace6c9 100644
> --- a/mm/memory-failure.c
> +++ b/mm/memory-failure.c
> @@ -1340,9 +1340,11 @@ static inline bool HWPoisonHandlable(struct page *page, unsigned long flags)
>  	if (PageSlab(page))
>  		return false;
> 
> +#ifdef CONFIG_MEMORY_FAILURE_MIGRATION
>  	/* Soft offline could migrate movable_ops pages */
>  	if ((flags & MF_SOFT_OFFLINE) && page_has_movable_ops(page))
>  		return true;
> +#endif
> 
>  	return PageLRU(page) || is_free_buddy_page(page);
>  }
> @@ -2028,7 +2030,8 @@ static int get_huge_page_for_hwpoison(unsigned long pfn, int flags,
>  		count_increased = true;
>  	} else if (folio_test_hugetlb_freed(folio)) {
>  		ret = MF_HUGETLB_FREED;
> -	} else if (folio_test_hugetlb_migratable(folio)) {
> +	} else if (IS_ENABLED(CONFIG_MEMORY_FAILURE_MIGRATION) &&
> +		   folio_test_hugetlb_migratable(folio)) {
>  		if (folio_try_get(folio)) {
>  			ret = MF_HUGETLB_IN_USED;
>  			count_increased = true;
> @@ -2051,10 +2054,12 @@ static int get_huge_page_for_hwpoison(unsigned long pfn, int flags,
>  	 * Clearing hugetlb_migratable for hwpoisoned hugepages to prevent them
>  	 * from being migrated by memory hotremove.
>  	 */
> +#ifdef CONFIG_MEMORY_FAILURE_MIGRATION
>  	if (count_increased && folio_test_hugetlb_migratable(folio)) {
>  		folio_clear_hugetlb_migratable(folio);
>  		*migratable_cleared = true;
>  	}
> +#endif
> 
>  	spin_unlock_irq(&hugetlb_lock);
>  	return ret;
> @@ -2903,7 +2908,7 @@ static int soft_offline_in_use_page(struct page *page)
>  		return 0;
>  	}
> 
> -	isolated = isolate_folio_to_list(folio, &pagelist);
> +	isolated = IS_ENABLED(CONFIG_MEMORY_FAILURE_MIGRATION) && isolate_folio_to_list(folio, &pagelist);
> 
>  	/*
>  	 * If we succeed to isolate the folio, we grabbed another refcount on
> ```
> 
> Would this be a better way?

Why add a new config option?

-- 
Cheers,

David