Re: [PATCH 04/25] mm/fbatch: lru bit set, no extra ref, while folio on per-cpu fbatch

[Hugh Dickins]

On Mon, 24 Aug 2026, Hugh Dickins wrote:
...
> --- a/include/linux/mm_inline.h
> +++ b/include/linux/mm_inline.h
> @@ -343,6 +343,23 @@ static inline void folio_migrate_refs(struct folio *new, const struct folio *old
>  }
>  #endif /* CONFIG_LRU_GEN */
>  
> +enum {
> +	LRU_NEXT_NEVER_TAIL = 0,	/* Used by a tail's compound_head */
> +	LRU_NEXT_BATCHED = 1,		/* Not used by any aligned pointer */
> +	NR_LRU_NEXT_FLAGS
> +};
> +
> +static __always_inline
> +bool lru_add_del_folio(struct folio *folio)
> +{
> +	/* BUG_ON(folio_test_lru(folio)); */

That was a good statement of intent, but clearly I never booted
a system with that BUG_ON uncommented: it actually needs to say
	/* BUG_ON(folio_test_lru(folio) && folio_ref_count(folio)); */
because once the refcount is 0, __page_cache_release() chooses to
lruvec_del_folio() before non-atomically clearing lru flags.
To be fixed when respun.

> +	if (!(folio->lru_next & BIT(LRU_NEXT_BATCHED)))
> +		return false;
> +	folio->lru.next = LIST_POISON1;
> +	/* BUG_ON(folio->lru_next & BIT(LRU_NEXT_BATCHED)); */
> +	return true;
> +}
...