Re: [PATCH v4 5/5] mm: filter out swapin error entry in shmem mapping

[HORIGUCHI NAOYA(堀口 直也)]

On Thu, May 19, 2022 at 08:50:30PM +0800, Miaohe Lin wrote:
> There might be swapin error entries in shmem mapping. Filter them out to
> avoid "Bad swap file entry" complaint.
> 
> Signed-off-by: Miaohe Lin <linmiaohe@xxxxxxxxxx>
> ---
>  mm/madvise.c    | 5 ++++-
>  mm/swap_state.c | 3 +++
>  2 files changed, 7 insertions(+), 1 deletion(-)
> 
> diff --git a/mm/madvise.c b/mm/madvise.c
> index a42165bc4735..31582b6ff551 100644
> --- a/mm/madvise.c
> +++ b/mm/madvise.c
> @@ -248,10 +248,13 @@ static void force_shm_swapin_readahead(struct vm_area_struct *vma,
>  
>  		if (!xa_is_value(page))
>  			continue;
> +		swap = radix_to_swp_entry(page);
> +		/* There might be swapin error entries in shmem mapping. */
> +		if (non_swap_entry(swap))
> +			continue;

The inline comment mentions swapin error entries but other types of
non-swap entries should be skipped by this check, which is helpful too.
So I'm fine with the change, thank you.

Reviewed-by: Naoya Horiguchi <naoya.horiguchi@xxxxxxx>

>  		xas_pause(&xas);
>  		rcu_read_unlock();
>  
> -		swap = radix_to_swp_entry(page);
>  		page = read_swap_cache_async(swap, GFP_HIGHUSER_MOVABLE,
>  					     NULL, 0, false, &splug);
>  		if (page)
> diff --git a/mm/swap_state.c b/mm/swap_state.c
> index b9e4ed2e90bf..778d57d2d92d 100644
> --- a/mm/swap_state.c
> +++ b/mm/swap_state.c
> @@ -410,6 +410,9 @@ struct page *find_get_incore_page(struct address_space *mapping, pgoff_t index)
>  		return NULL;
>  
>  	swp = radix_to_swp_entry(page);
> +	/* There might be swapin error entries in shmem mapping. */
> +	if (non_swap_entry(swp))
> +		return NULL;
>  	/* Prevent swapoff from happening to us */
>  	si = get_swap_device(swp);
>  	if (!si)
> -- 
> 2.23.0