Re: [PATCH bpf-next v2 2/6] mm, bpf: Introduce free_pages_nolock()

[Vlastimil Babka <vbabka@suse.cz>]

On 12/10/24 03:39, Alexei Starovoitov wrote:
> From: Alexei Starovoitov <ast@kernel.org>
> 
> Introduce free_pages_nolock() that can free a page without taking locks.
> It relies on trylock only and can be called from any context.
> 
> Signed-off-by: Alexei Starovoitov <ast@kernel.org>

<snip>

> +/* Can be called while holding raw_spin_lock or from IRQ. RCU must be watching. */
> +void free_pages_nolock(struct page *page, unsigned int order)

What does "RCU must be watching" mean and why?

> +{
> +	int head = PageHead(page);
> +	struct alloc_tag *tag = pgalloc_tag_get(page);
> +
> +	if (put_page_testzero(page)) {
> +		__free_unref_page(page, order, FPI_TRYLOCK);
> +	} else if (!head) {
> +		pgalloc_tag_sub_pages(tag, (1 << order) - 1);
> +		while (order-- > 0)
> +			__free_unref_page(page + (1 << order), order, FPI_TRYLOCK);

Not your fault for trying to support everything __free_pages did,
specifically order > 0 pages that are not compound and thus needing this
extra !head branch. We'd love to get rid of that eventually in
__free_pages(), but at least I think we don't need to support it in a new
API and instead any users switching to it should know it's not supported.
I suppose BFP doesn't need that, right?
Maybe if the function was taking a folio instead of page, it would be the
best as that has to be order-0 or compound by definition. It also wouldn't
need the order parameter. What do you think, Matthew?

> +	}
> +}
> +
>  void free_pages(unsigned long addr, unsigned int order)
>  {
>  	if (addr != 0) {