Re: [PATCH net-next v8 06/17] net: page pool: add helper creating area from pages

[Jakub Kicinski <kuba@kernel.org>]

On Wed,  4 Dec 2024 09:21:45 -0800 David Wei wrote:
> From: Pavel Begunkov <asml.silence@gmail.com>
> 
> Add a helper that takes an array of pages and initialises passed in
> memory provider's area with them, where each net_iov takes one page.
> It's also responsible for setting up dma mappings.
> 
> We keep it in page_pool.c not to leak netmem details to outside
> providers like io_uring, which don't have access to netmem_priv.h
> and other private helpers.

User space will likely give us hugepages. Feels a bit wasteful to map
and manage them 4k at a time. But okay, we can optimize this later.

> diff --git a/include/net/page_pool/memory_provider.h b/include/net/page_pool/memory_provider.h
> new file mode 100644
> index 000000000000..83d7eec0058d
> --- /dev/null
> +++ b/include/net/page_pool/memory_provider.h
> @@ -0,0 +1,10 @@

nit: missing SPDX

> +#ifndef _NET_PAGE_POOL_MEMORY_PROVIDER_H
> +#define _NET_PAGE_POOL_MEMORY_PROVIDER_H
> +
> +int page_pool_mp_init_paged_area(struct page_pool *pool,
> +				struct net_iov_area *area,
> +				struct page **pages);
> +void page_pool_mp_release_area(struct page_pool *pool,
> +				struct net_iov_area *area);
> +
> +#endif

> +static void page_pool_release_page_dma(struct page_pool *pool,
> +				       netmem_ref netmem)
> +{
> +	__page_pool_release_page_dma(pool, netmem);

I'm guessing this is to save text? Because
__page_pool_release_page_dma() is always_inline? 
Maybe add a comment?

> +}
> +
> +int page_pool_mp_init_paged_area(struct page_pool *pool,
> +				 struct net_iov_area *area,
> +				 struct page **pages)
> +{
> +	struct net_iov *niov;
> +	netmem_ref netmem;
> +	int i, ret = 0;
> +
> +	if (!pool->dma_map)
> +		return -EOPNOTSUPP;
> +
> +	for (i = 0; i < area->num_niovs; i++) {
> +		niov = &area->niovs[i];
> +		netmem = net_iov_to_netmem(niov);
> +
> +		page_pool_set_pp_info(pool, netmem);

Maybe move setting pp down, after we successfully mapped. Technically
it's not a bug to leave it set on netmem, but it would be on a page
struct.

> +		if (!page_pool_dma_map_page(pool, netmem, pages[i])) {
> +			ret = -EINVAL;
> +			goto err_unmap_dma;
> +		}
> +	}
> +	return 0;
> +
> +err_unmap_dma:
> +	while (i--) {
> +		netmem = net_iov_to_netmem(&area->niovs[i]);
> +		page_pool_release_page_dma(pool, netmem);
> +	}
> +	return ret;
> +}
> +
> +void page_pool_mp_release_area(struct page_pool *pool,
> +			       struct net_iov_area *area)
> +{
> +	int i;
> +
> +	if (!pool->dma_map)
> +		return;
> +
> +	for (i = 0; i < area->num_niovs; i++) {
> +		struct net_iov *niov = &area->niovs[i];
> +
> +		page_pool_release_page_dma(pool, net_iov_to_netmem(niov));
> +	}
> +}