Re: [PATCH 5/8] migrate_pages: batch _unmap and _move

[Andrew Morton]

On Tue, 27 Dec 2022 08:28:56 +0800 Huang Ying <ying.huang@xxxxxxxxx> wrote:

> In this patch the _unmap and _move stage of the folio migration is
> batched.  That for, previously, it is,
> 
>   for each folio
>     _unmap()
>     _move()
> 
> Now, it is,
> 
>   for each folio
>     _unmap()
>   for each folio
>     _move()
> 
> Based on this, we can batch the TLB flushing and use some hardware
> accelerator to copy folios between batched _unmap and batched _move
> stages.
> 
> ...
>
> --- a/mm/migrate.c
> +++ b/mm/migrate.c
> @@ -1027,8 +1027,32 @@ static void __migrate_folio_extract(struct folio *dst,
>  	dst->private = NULL;
>  }
>  
> +static void migrate_folio_undo_src(struct folio *src,
> +				   int page_was_mapped,
> +				   struct anon_vma *anon_vma,
> +				   struct list_head *ret)
> +{
> +	if (page_was_mapped)
> +		remove_migration_ptes(src, src, false);
> +	if (anon_vma)
> +		put_anon_vma(anon_vma);
> +	folio_unlock(src);
> +	list_move_tail(&src->lru, ret);
> +}
> +
> +static void migrate_folio_undo_dst(struct folio *dst,
> +				   free_page_t put_new_page,
> +				   unsigned long private)
> +{
> +	folio_unlock(dst);
> +	if (put_new_page)
> +		put_new_page(&dst->page, private);
> +	else
> +		folio_put(dst);
> +}

What do the above do?  Are they so obvious that no comments are needed?


>  static int __migrate_folio_unmap(struct folio *src, struct folio *dst,
> -				int force, enum migrate_mode mode)
> +				 int force, bool force_lock, enum migrate_mode mode)
>  {
>  	int rc = -EAGAIN;
>  	int page_was_mapped = 0;
> @@ -1055,6 +1079,11 @@ static int __migrate_folio_unmap(struct folio *src, struct folio *dst,
>  		if (current->flags & PF_MEMALLOC)
>  			goto out;
>  
> +		if (!force_lock) {
> +			rc = -EDEADLOCK;
> +			goto out;
> +		}

Please document the use of EDEADLOCK in this code.  What does it signify?

>  		folio_lock(src);
>  	}
>