net/mlx5e: Introduce RX Page-Reuse

Introduce a Page-Reuse mechanism in non-Striding RQ RX datapath.

A WQE (RX descriptor) buffer is a page, that in most cases was fully
wasted on a packet that is much smaller, requiring a new page for
the next round.

In this patch, we implement a page-reuse mechanism, that resembles a
`SW Striding RQ`.
We allow the WQE to reuse its allocated page as much as it could,
until the page is fully consumed.  In each round, the WQE is capable
of receiving packet of maximal size (MTU). Yet, upon the reception of
a packet, the WQE knows the actual packet size, and consumes the exact
amount of memory needed to build a linear SKB. Then, it updates the
buffer pointer within the page accordingly, for the next round.

Feature is mutually exclusive with XDP (packet-per-page)
and LRO (session size is a power of two, needs unused page).

Performance tests:
iperf tcp tests show huge gain:

--------------------------------------------
num streams | BW before | BW after | ratio |
          1 |      22.2 |     30.9 | 1.39x |
          8 |      64.2 |     93.6 | 1.46x |
         64 |      56.7 |     91.4 | 1.61x |
--------------------------------------------

Signed-off-by: Tariq Toukan <tariqt@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>

1 files changed, 20 insertions, 8 deletions

diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en_main.c b/drivers/net/ethernet/mellanox/mlx5/core/en_main.c
index 66173e5545ce..9f99f624004f 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en_main.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en_main.c
@@ -200,6 +200,7 @@ static void mlx5e_update_sw_counters(struct mlx5e_priv *priv)
 		s->rx_buff_alloc_err += rq_stats->buff_alloc_err;
 		s->rx_cqe_compress_blks += rq_stats->cqe_compress_blks;
 		s->rx_cqe_compress_pkts += rq_stats->cqe_compress_pkts;
+		s->rx_page_reuse  += rq_stats->page_reuse;
 		s->rx_cache_reuse += rq_stats->cache_reuse;
 		s->rx_cache_full  += rq_stats->cache_full;
 		s->rx_cache_empty += rq_stats->cache_empty;
@@ -550,7 +551,6 @@ static int mlx5e_alloc_rq(struct mlx5e_channel *c,
 	void *rqc = rqp->rqc;
 	void *rqc_wq = MLX5_ADDR_OF(rqc, rqc, wq);
 	u32 byte_count;
-	u32 frag_sz;
 	int npages;
 	int wq_sz;
 	int err;
@@ -614,9 +614,10 @@ static int mlx5e_alloc_rq(struct mlx5e_channel *c,
 			goto err_destroy_umr_mkey;
 		break;
 	default: /* MLX5_WQ_TYPE_LINKED_LIST */
-		rq->dma_info = kzalloc_node(wq_sz * sizeof(*rq->dma_info),
-					    GFP_KERNEL, cpu_to_node(c->cpu));
-		if (!rq->dma_info) {
+		rq->wqe.frag_info =
+			kzalloc_node(wq_sz * sizeof(*rq->wqe.frag_info),
+				     GFP_KERNEL, cpu_to_node(c->cpu));
+		if (!rq->wqe.frag_info) {
 			err = -ENOMEM;
 			goto err_rq_wq_destroy;
 		}
@@ -625,7 +626,7 @@ static int mlx5e_alloc_rq(struct mlx5e_channel *c,
 
 		rq->handle_rx_cqe = c->priv->profile->rx_handlers.handle_rx_cqe;
 		if (!rq->handle_rx_cqe) {
-			kfree(rq->dma_info);
+			kfree(rq->wqe.frag_info);
 			err = -EINVAL;
 			netdev_err(c->netdev, "RX handler of RQ is not set, err %d\n", err);
 			goto err_rq_wq_destroy;
@@ -634,11 +635,12 @@ static int mlx5e_alloc_rq(struct mlx5e_channel *c,
 		rq->buff.wqe_sz = params->lro_en  ?
 				params->lro_wqe_sz :
 				MLX5E_SW2HW_MTU(c->priv, c->netdev->mtu);
+		rq->wqe.page_reuse = !params->xdp_prog && !params->lro_en;
 		byte_count = rq->buff.wqe_sz;
 
 		/* calc the required page order */
-		frag_sz = MLX5_SKB_FRAG_SZ(rq->rx_headroom + byte_count);
-		npages = DIV_ROUND_UP(frag_sz, PAGE_SIZE);
+		rq->wqe.frag_sz = MLX5_SKB_FRAG_SZ(rq->rx_headroom + byte_count);
+		npages = DIV_ROUND_UP(rq->wqe.frag_sz, PAGE_SIZE);
 		rq->buff.page_order = order_base_2(npages);
 
 		byte_count |= MLX5_HW_START_PADDING;
@@ -683,7 +685,7 @@ static void mlx5e_free_rq(struct mlx5e_rq *rq)
 		mlx5_core_destroy_mkey(rq->mdev, &rq->umr_mkey);
 		break;
 	default: /* MLX5_WQ_TYPE_LINKED_LIST */
-		kfree(rq->dma_info);
+		kfree(rq->wqe.frag_info);
 	}
 
 	for (i = rq->page_cache.head; i != rq->page_cache.tail;
@@ -865,6 +867,16 @@ static void mlx5e_free_rx_descs(struct mlx5e_rq *rq)
 		mlx5_wq_ll_pop(&rq->wq, wqe_ix_be,
 			       &wqe->next.next_wqe_index);
 	}
+
+	if (rq->wq_type == MLX5_WQ_TYPE_LINKED_LIST && rq->wqe.page_reuse) {
+		/* Clean outstanding pages on handled WQEs that decided to do page-reuse,
+		 * but yet to be re-posted.
+		 */
+		int wq_sz = mlx5_wq_ll_get_size(&rq->wq);
+
+		for (wqe_ix = 0; wqe_ix < wq_sz; wqe_ix++)
+			rq->dealloc_wqe(rq, wqe_ix);
+	}
 }
 
 static int mlx5e_open_rq(struct mlx5e_channel *c,