Re: [PATCH 4/5] mm: add largest_zero_folio() routine

[Dave Hansen <dave.hansen@intel.com>]

On 8/4/25 05:13, Pankaj Raghav (Samsung) wrote:
> From: Pankaj Raghav <p.raghav@samsung.com>
> 
> Add largest_zero_folio() routine so that huge_zero_folio can be
> used directly when CONFIG_STATIC_HUGE_ZERO_FOLIO is enabled. This will
> return ZERO_PAGE folio if CONFIG_STATIC_HUGE_ZERO_FOLIO is disabled or
> if we failed to allocate a huge_zero_folio.

This changelog is telling a lot of the "what" but none of the "why".

The existing huge zero folio API is for users that have an mm. This is
*only* for folks that want a huge zero folio but don't have an mm.
That's _why_ this function is needed. It's in this series because there
was no way to get a huge zero folio before the permanent one was
introduced in this series.

Can we please get some of that into the function comment and changelog?
It's critical.

> diff --git a/include/linux/huge_mm.h b/include/linux/huge_mm.h
> index 78ebceb61d0e..c44a6736704b 100644
> --- a/include/linux/huge_mm.h
> +++ b/include/linux/huge_mm.h
> @@ -716,4 +716,21 @@ static inline int split_folio_to_order(struct folio *folio, int new_order)
>  	return split_folio_to_list_to_order(folio, NULL, new_order);
>  }
>  
> +/*
> + * largest_zero_folio - Get the largest zero size folio available
> + *
> + * This function will return huge_zero_folio if CONFIG_STATIC_HUGE_ZERO_FOLIO
> + * is enabled. Otherwise, a ZERO_PAGE folio is returned.
> + *
> + * Deduce the size of the folio with folio_size instead of assuming the
> + * folio size.
> + */

This comment needs to get fleshed out. It at _least_ needs to say
something along the lines of:

	Use this when a huge zero folio is needed but there is no mm
	lifetime to tie it to. Basically, use this when you can't use
	mm_get_huge_zero_folio().

> +static inline struct folio *largest_zero_folio(void)
> +{
> +	struct folio *folio = get_static_huge_zero_folio();
> +
> +	if (folio)
> +		return folio;

There needs to be a newline in here.

> +	return page_folio(ZERO_PAGE(0));
> +}
>  #endif /* _LINUX_HUGE_MM_H */

The code is fine, but the changelog and comments need quite a bit of work.