Re: [PATCH v1] mm/vmscan: check the return value of migrate_pages()

[Baolin Wang]

Hi,

On 9/17/2022 10:12 AM, Li Zhong wrote:
> Check the return value of migrate_pages() to check error exeuction.
>
> Signed-off-by: Li Zhong <floridsleeves@xxxxxxxxx>
> ---
>   mm/vmscan.c | 6 +++++-
>   1 file changed, 5 insertions(+), 1 deletion(-)
>
> diff --git a/mm/vmscan.c b/mm/vmscan.c
> index b2b1431352dc..50aaaa9377e2 100644
> --- a/mm/vmscan.c
> +++ b/mm/vmscan.c
> @@ -1547,6 +1547,7 @@ static unsigned int demote_page_list(struct list_head *demote_pages,
>   {
>   	int target_nid = next_demotion_node(pgdat->node_id);
>   	unsigned int nr_succeeded;
> +	int ret;
>   
>   	if (list_empty(demote_pages))
>   		return 0;
> @@ -1555,10 +1556,13 @@ static unsigned int demote_page_list(struct list_head *demote_pages,
>   		return 0;
>   
>   	/* Demotion ignores all cpuset and mempolicy settings */
> -	migrate_pages(demote_pages, alloc_demote_page, NULL,
> +	ret = migrate_pages(demote_pages, alloc_demote_page, NULL,
>   			    target_nid, MIGRATE_ASYNC, MR_DEMOTION,
>   			    &nr_succeeded);
>   
> +	if (ret)
> +		putback_movable_pages(&pagelist);

You can not just put them back into the original lru list, instead the 
caller has handled the failed-to-demote pages, which will try to reclaim 
them.

In addition, the code cannot even be built successfully.