Re: [PATCH v4 net-next 2/5] net: page_pool: add bulk support for ptr_ring

[Jesper Dangaard Brouer <brouer@redhat.com>]

On Fri,  6 Nov 2020 19:19:08 +0100
Lorenzo Bianconi <lorenzo@kernel.org> wrote:

> Introduce the capability to batch page_pool ptr_ring refill since it is
> usually run inside the driver NAPI tx completion loop.
> 
> Suggested-by: Jesper Dangaard Brouer <brouer@redhat.com>
> Signed-off-by: Lorenzo Bianconi <lorenzo@kernel.org>
> ---
>  include/net/page_pool.h | 26 ++++++++++++++++
>  net/core/page_pool.c    | 66 ++++++++++++++++++++++++++++++++++-------
>  net/core/xdp.c          |  9 ++----
>  3 files changed, 84 insertions(+), 17 deletions(-)

My own suggestion to create __page_pool_put_page() for code sharing,
cause a performance regression, because compiler chooses not to inline
the function.  This cause unnecessary code for in_serving_softirq() and
function call overhead.

I'm currently testing with __always_inline and likely/unlikely
annotation to get compiler to layout code better for I-cache.  This
shows improvements in my benchmarks. Details in[1].

[1] https://github.com/xdp-project/xdp-project/blob/master/areas/mem/xdp_bulk_return01.org

> diff --git a/net/core/page_pool.c b/net/core/page_pool.c
> index ef98372facf6..31dac2ad4a1f 100644
> --- a/net/core/page_pool.c
> +++ b/net/core/page_pool.c
[...]
> @@ -362,8 +364,9 @@ static bool pool_page_reusable(struct page_pool *pool, struct page *page)
>   * If the page refcnt != 1, then the page will be returned to memory
>   * subsystem.
>   */
> -void page_pool_put_page(struct page_pool *pool, struct page *page,
> -			unsigned int dma_sync_size, bool allow_direct)
> +static struct page *
> +__page_pool_put_page(struct page_pool *pool, struct page *page,
> +		     unsigned int dma_sync_size, bool allow_direct)
>  {
>  	/* This allocator is optimized for the XDP mode that uses
>  	 * one-frame-per-page, but have fallbacks that act like the
> @@ -379,15 +382,12 @@ void page_pool_put_page(struct page_pool *pool, struct page *page,
>  			page_pool_dma_sync_for_device(pool, page,
>  						      dma_sync_size);
>  
> -		if (allow_direct && in_serving_softirq())
> -			if (page_pool_recycle_in_cache(page, pool))
> -				return;
> +		if (allow_direct && in_serving_softirq() &&
> +		    page_pool_recycle_in_cache(page, pool))
> +			return NULL;
>  
> -		if (!page_pool_recycle_in_ring(pool, page)) {
> -			/* Cache full, fallback to free pages */
> -			page_pool_return_page(pool, page);
> -		}
> -		return;
> +		/* page is candidate for bulking */
> +		return page;
>  	}
>  	/* Fallback/non-XDP mode: API user have elevated refcnt.
>  	 *
> @@ -405,9 +405,55 @@ void page_pool_put_page(struct page_pool *pool, struct page *page,
>  	/* Do not replace this with page_pool_return_page() */
>  	page_pool_release_page(pool, page);
>  	put_page(page);
> +
> +	return NULL;
> +}
> +
> +void page_pool_put_page(struct page_pool *pool, struct page *page,
> +			unsigned int dma_sync_size, bool allow_direct)
> +{
> +	page = __page_pool_put_page(pool, page, dma_sync_size, allow_direct);
> +	if (page && !page_pool_recycle_in_ring(pool, page)) {
> +		/* Cache full, fallback to free pages */
> +		page_pool_return_page(pool, page);
> +	}
>  }
>  EXPORT_SYMBOL(page_pool_put_page);
>  
> +/* Caller must not use data area after call, as this function overwrites it */
> +void page_pool_put_page_bulk(struct page_pool *pool, void **data,
> +			     int count)
> +{
> +	int i, bulk_len = 0, pa_len = 0;
> +
> +	for (i = 0; i < count; i++) {
> +		struct page *page = virt_to_head_page(data[i]);
> +
> +		page = __page_pool_put_page(pool, page, -1, false);
> +		/* Approved for bulk recycling in ptr_ring cache */
> +		if (page)
> +			data[bulk_len++] = page;
> +	}
> +
> +	if (!bulk_len)
> +		return;
> +
> +	/* Bulk producer into ptr_ring page_pool cache */
> +	page_pool_ring_lock(pool);
> +	for (i = 0; i < bulk_len; i++) {
> +		if (__ptr_ring_produce(&pool->ring, data[i]))
> +			data[pa_len++] = data[i];
> +	}
> +	page_pool_ring_unlock(pool);
> +
> +	/* ptr_ring cache full, free pages outside producer lock since
> +	 * put_page() with refcnt == 1 can be an expensive operation
> +	 */
> +	for (i = 0; i < pa_len; i++)
> +		page_pool_return_page(pool, data[i]);
> +}
> +EXPORT_SYMBOL(page_pool_put_page_bulk);


-- 
Best regards,
  Jesper Dangaard Brouer
  MSc.CS, Principal Kernel Engineer at Red Hat
  LinkedIn: http://www.linkedin.com/in/brouer