xdp: introduce bulking for page_pool tx return path

XDP bulk APIs introduce a defer/flush mechanism to return pages belonging to the same xdp_mem_allocator object (identified via the mem.id field) in bulk to optimize I-cache and D-cache since xdp_return_frame is usually run inside the driver NAPI tx completion loop.
The bulk queue size is set to 16 to be aligned to how XDP_REDIRECT bulking works. The bulk is flushed when it is full or when mem.id changes.
xdp_frame_bulk is usually stored/allocated on the function call-stack to avoid locking penalties.
Current implementation considers only page_pool memory model.

Suggested-by Jesper Dangaard Brouer <[email protected]>

Signed-off-by Lorenzo Bianconi <[email protected]>

---
 include/net/xdp.h | 11 ++++++++-
 net/core/xdp.c    | 61 +++++++++++++++++++++++++++++++++++++++++++++++
 2 files changed, 71 insertions(+), 1 deletion(-)

net/core/xdp.c +20 -0

+	xa = bq->xa;
+	if (unlikely(!xa)) {
+		xa = rhashtable_lookup(mem_id_ht, &mem->id, mem_id_rht_params);
+		bq->count = 0;
+		bq->xa = xa;
+	}
+
+	if (bq->count == XDP_BULK_QUEUE_SIZE)
+		xdp_flush_frame_bulk(bq);
+
+	if (mem->id != xa->mem.id) {
+		xdp_flush_frame_bulk(bq);
+		bq->xa = rhashtable_lookup(mem_id_ht, &mem->id, mem_id_rht_params);
+	}
+
+	bq->q[bq->count++] = xdpf->data;
+
+	rcu_read_unlock();
+}
+EXPORT_SYMBOL_GPL(xdp_return_frame_bulk);

net/core/page_pool.c +5 -5

index ef98372facf6..31dac2ad4a1f 100644
@@ -11,6 +11,8 @@
 #include <linux/device.h>
 
 #include <net/page_pool.h>
+#include <net/xdp.h>
+
 #include <linux/dma-direction.h>
 #include <linux/dma-mapping.h>
 #include <linux/page-flags.h>

@@ -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;

net/core/page_pool.c +36 -0

+		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];

On Fri, 6 Nov 2020 19:19:08 +0100
Lorenzo Bianconi <[email protected]> wrote:

I don't like this comment. This function is also used by non-bulking case. Reaching this point means the page is ready or fulfill the requirement for being recycled into the ptr_ring.

I suggest (as before):

		/* Page found as candidate for recycling */

Quoted 47 lines

net/core/page_pool.c +36 -0

+		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];

net/core/page_pool.c void page_pool_put_page(struct page_pool *pool, struct page *page, · line 455

Quoted 10 lines

net/core/page_pool.c +10 -0

+	}
+	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);

Rest looks okay :-)

Signature · 4 lines

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

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 <[email protected]>

Signed-off-by Lorenzo Bianconi <[email protected]>

---
 include/net/page_pool.h | 26 ++++++++++++++++
 net/core/page_pool.c    | 66 ++++++++++++++++++++++++++++++++++-------
 net/core/xdp.c          |  9 ++----
 3 files changed, 84 insertions(+), 17 deletions(-)

net/core/page_pool.c +0 -0

index ef98372facf6..31dac2ad4a1f 100644

net/core/page_pool.c +53 -10

@@ -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);

Convert mlx5 driver to xdp_return_frame_bulk APIs.

XDP_REDIRECT (upstream codepath): 8.5Mpps XDP_REDIRECT (upstream codepath + bulking APIs): 10.1Mpps

Signed-off-by Lorenzo Bianconi <[email protected]>

---
 drivers/net/ethernet/mellanox/mlx5/core/en/xdp.c | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)

drivers/net/ethernet/mellanox/mlx5/core/en/xdp.c +4 -1

index ae90d533a350..5fdfbf390d5c 100644
@@ -369,8 +369,10 @@ static void mlx5e_free_xdpsq_desc(struct mlx5e_xdpsq *sq,
 				  bool recycle)
 {
 	struct mlx5e_xdp_info_fifo *xdpi_fifo = &sq->db.xdpi_fifo;
+	struct xdp_frame_bulk bq;
 	u16 i;
 
+	bq.xa = NULL;
 	for (i = 0; i < wi->num_pkts; i++) {
 		struct mlx5e_xdp_info xdpi = mlx5e_xdpi_fifo_pop(xdpi_fifo);
 

@@ -379,7 +381,7 @@ static void mlx5e_free_xdpsq_desc(struct mlx5e_xdpsq *sq,
 			/* XDP_TX from the XSK RQ and XDP_REDIRECT */
 			dma_unmap_single(sq->pdev, xdpi.frame.dma_addr,
 					 xdpi.frame.xdpf->len, DMA_TO_DEVICE);
-			xdp_return_frame(xdpi.frame.xdpf);
+			xdp_return_frame_bulk(xdpi.frame.xdpf, &bq);
 			break;
 		case MLX5E_XDP_XMIT_MODE_PAGE:
 			/* XDP_TX from the regular RQ */

@@ -393,6 +395,7 @@ static void mlx5e_free_xdpsq_desc(struct mlx5e_xdpsq *sq,
 			WARN_ON_ONCE(true);
 		}
 	}
+	xdp_flush_frame_bulk(&bq);
 }
 
 bool mlx5e_poll_xdpsq_cq(struct mlx5e_cq *cq)