Re: [PATCH] mm, compaction: avoid isolating pinned tmpfs pages

[Yu Xu <xuyu@linux.alibaba.com>]

On 4/28/21 10:00 PM, Xu Yu wrote:
> This makes pinned tmpfs pages bail out early in the process of page
> migration, like what pinned anonymous pages do.
> 
> Signed-off-by: Xu Yu <xuyu@linux.alibaba.com>
> ---
>   mm/compaction.c | 8 ++++----
>   1 file changed, 4 insertions(+), 4 deletions(-)
> 
> diff --git a/mm/compaction.c b/mm/compaction.c
> index e04f4476e68e..78c3b992a1c9 100644
> --- a/mm/compaction.c
> +++ b/mm/compaction.c
> @@ -964,11 +964,11 @@ isolate_migratepages_block(struct compact_control *cc, unsigned long low_pfn,
>   		}
>   
>   		/*
> -		 * Migration will fail if an anonymous page is pinned in memory,
> -		 * so avoid taking lru_lock and isolating it unnecessarily in an
> -		 * admittedly racy check.
> +		 * Migration will fail if an anonymous or tmpfs page is pinned
> +		 * in memory, so avoid taking lru_lock and isolating it
> +		 * unnecessarily in an admittedly racy check.
>   		 */
> -		if (!page_mapping(page) &&
> +		if (!page_is_file_lru(page) &&
>   		    page_count(page) > page_mapcount(page))
Sorry, this patch is flawed, since tmpfs page cache takes an extra page
count.

Please ignore this patch.

>   			goto isolate_fail;
>   
> 

-- 
Thanks,
Yu