Re: [PATCH 08/16] mm/page_alloc: add missing is_migrate_isolate() check in set_page_guard()

[Oscar Salvador]

On Fri, Sep 09, 2022 at 05:24:43PM +0800, Miaohe Lin wrote:
> In MIGRATE_ISOLATE case, zone freepage state shouldn't be modified as
> caller will take care of it. Add missing is_migrate_isolate() here to
> avoid possible unbalanced freepage state.
> 
> Signed-off-by: Miaohe Lin <linmiaohe@xxxxxxxxxx>

Ok, I gave it some thought and I guess it's safe to assume that this would happen if
someone isolates the block, and then we face an MCE failure/soft-offline on a page
within that block.

take_page_off_buddy
 break_down_buddy_pages
  set_page_guard

will trigger __mod_zone_freepage_state(), which already had been triggered back
when the block was isolated.

I think the changelog could grow fatter to better explain the issue.

Reviewed-by: Oscar Salvador <osalvador@xxxxxxx>

> ---
>  mm/page_alloc.c | 3 ++-
>  1 file changed, 2 insertions(+), 1 deletion(-)
> 
> diff --git a/mm/page_alloc.c b/mm/page_alloc.c
> index a35ef385d906..94baf33da865 100644
> --- a/mm/page_alloc.c
> +++ b/mm/page_alloc.c
> @@ -873,7 +873,8 @@ static inline bool set_page_guard(struct zone *zone, struct page *page,
>  	INIT_LIST_HEAD(&page->buddy_list);
>  	set_page_private(page, order);
>  	/* Guard pages are not available for any usage */
> -	__mod_zone_freepage_state(zone, -(1 << order), migratetype);
> +	if (!is_migrate_isolate(migratetype))
> +		__mod_zone_freepage_state(zone, -(1 << order), migratetype);
>  
>  	return true;
>  }
> -- 
> 2.23.0
> 

-- 
Oscar Salvador
SUSE Labs