mlx4: add rx_alloc_pages counter in ethtool -S

This new counter tracks number of pages that we allocated for one port.

lpaa24:~# ethtool -S eth0 | egrep 'rx_alloc_pages|rx_packets'
     rx_packets: 306755183
     rx_alloc_pages: 932897

Signed-off-by: Eric Dumazet <edumazet@google.com>
Acked-by: Tariq Toukan <tariqt@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

drivers/net/ethernet/mellanox/mlx4/en_ethtool.c

@@ -117,7 +117,7 @@ static const char main_strings[][ETH_GSTRING_LEN] = {
 	/* port statistics */
 	"tso_packets",
 	"xmit_more",
-	"queue_stopped", "wake_queue", "tx_timeout", "rx_alloc_failed",
+	"queue_stopped", "wake_queue", "tx_timeout", "rx_alloc_pages",
 	"rx_csum_good", "rx_csum_none", "rx_csum_complete", "tx_chksum_offload",
 
 	/* pf statistics */

drivers/net/ethernet/mellanox/mlx4/en_port.c

@@ -213,6 +213,7 @@ int mlx4_en_DUMP_ETH_STATS(struct mlx4_en_dev *mdev, u8 port, u8 reset)
 	priv->port_stats.rx_chksum_good = 0;
 	priv->port_stats.rx_chksum_none = 0;
 	priv->port_stats.rx_chksum_complete = 0;
+	priv->port_stats.rx_alloc_pages = 0;
 	priv->xdp_stats.rx_xdp_drop    = 0;
 	priv->xdp_stats.rx_xdp_tx      = 0;
 	priv->xdp_stats.rx_xdp_tx_full = 0;
@@ -223,6 +224,7 @@ int mlx4_en_DUMP_ETH_STATS(struct mlx4_en_dev *mdev, u8 port, u8 reset)
 		priv->port_stats.rx_chksum_good += READ_ONCE(ring->csum_ok);
 		priv->port_stats.rx_chksum_none += READ_ONCE(ring->csum_none);
 		priv->port_stats.rx_chksum_complete += READ_ONCE(ring->csum_complete);
+		priv->port_stats.rx_alloc_pages += READ_ONCE(ring->rx_alloc_pages);
 		priv->xdp_stats.rx_xdp_drop	+= READ_ONCE(ring->xdp_drop);
 		priv->xdp_stats.rx_xdp_tx	+= READ_ONCE(ring->xdp_tx);
 		priv->xdp_stats.rx_xdp_tx_full	+= READ_ONCE(ring->xdp_tx_full);

drivers/net/ethernet/mellanox/mlx4/en_rx.c

@@ -72,6 +72,7 @@ static int mlx4_alloc_page(struct mlx4_en_priv *priv,
 }
 
 static int mlx4_en_alloc_frags(struct mlx4_en_priv *priv,
+			       struct mlx4_en_rx_ring *ring,
 			       struct mlx4_en_rx_desc *rx_desc,
 			       struct mlx4_en_rx_alloc *frags,
 			       gfp_t gfp)
@@ -79,8 +80,11 @@ static int mlx4_en_alloc_frags(struct mlx4_en_priv *priv,
 	int i;
 
 	for (i = 0; i < priv->num_frags; i++, frags++) {
-		if (!frags->page && mlx4_alloc_page(priv, frags, gfp))
-			return -ENOMEM;
+		if (!frags->page) {
+			if (mlx4_alloc_page(priv, frags, gfp))
+				return -ENOMEM;
+			ring->rx_alloc_pages++;
+		}
 		rx_desc->data[i].addr = cpu_to_be64(frags->dma +
 						    frags->page_offset);
 	}
@@ -133,7 +137,6 @@ static int mlx4_en_prepare_rx_desc(struct mlx4_en_priv *priv,
 	struct mlx4_en_rx_desc *rx_desc = ring->buf + (index * ring->stride);
 	struct mlx4_en_rx_alloc *frags = ring->rx_info +
 					(index << priv->log_rx_info);
-
 	if (ring->page_cache.index > 0) {
 		/* XDP uses a single page per frame */
 		if (!frags->page) {
@@ -147,7 +150,7 @@ static int mlx4_en_prepare_rx_desc(struct mlx4_en_priv *priv,
 		return 0;
 	}
 
-	return mlx4_en_alloc_frags(priv, rx_desc, frags, gfp);
+	return mlx4_en_alloc_frags(priv, ring, rx_desc, frags, gfp);
 }
 
 static bool mlx4_en_is_ring_empty(const struct mlx4_en_rx_ring *ring)

drivers/net/ethernet/mellanox/mlx4/mlx4_en.h

@@ -346,6 +346,7 @@ struct mlx4_en_rx_ring {
 	unsigned long csum_ok;
 	unsigned long csum_none;
 	unsigned long csum_complete;
+	unsigned long rx_alloc_pages;
 	unsigned long xdp_drop;
 	unsigned long xdp_tx;
 	unsigned long xdp_tx_full;

drivers/net/ethernet/mellanox/mlx4/mlx4_stats.h

@@ -37,7 +37,7 @@ struct mlx4_en_port_stats {
 	unsigned long queue_stopped;
 	unsigned long wake_queue;
 	unsigned long tx_timeout;
-	unsigned long rx_alloc_failed;
+	unsigned long rx_alloc_pages;
 	unsigned long rx_chksum_good;
 	unsigned long rx_chksum_none;
 	unsigned long rx_chksum_complete;