Merge branch 'stmmac-per-queue-stats'

Vijayakannan Ayyathurai says:

====================
net: stmmac: Add ethtool per-queue statistic

Adding generic ethtool per-queue statistic framework to display the
statistics for each rx/tx queue. In future, users can avail it to add
more per-queue specific counters. Number of rx/tx queues displayed is
depending on the available rx/tx queues in that particular MAC config
and this number is limited up to the MTL_MAX_{RX|TX}_QUEUES defined
in the driver.

Ethtool per-queue statistic display will look like below, when users
start adding more counters.

Example - 1:
 q0_tx_statA:
 q0_tx_statB:
 q0_tx_statC:
 |
 q0_tx_statX:
 .
 .
 .
 qMAX_tx_statA:
 qMAX_tx_statB:
 qMAX_tx_statC:
 |
 qMAX_tx_statX:

 q0_rx_statA:
 q0_rx_statB:
 q0_rx_statC:
 |
 q0_rx_statX:
 .
 .
 .
 qMAX_rx_statA:
 qMAX_rx_statB:
 qMAX_rx_statC:
 |
 qMAX_rx_statX:

Example - 2: Ping test using the tx queue 3.

$ tc qdisc add dev enp0s30f4 root mqprio num_tc 2 map 1 0 0 0 0 0 0 0
 0 0 0 0 0 0 0 0 queues 3@0 1@3 hw 0

Statistic before ping:
---------------------
$ ethtool -S enp0s30f4

[ snip ]
     q3_tx_pkt_n: 7916
     q3_tx_irq_n: 316
[ snip ]

$ cat /proc/interrupts

[ snip ]
 143:          0          0          0        316          0          0

         0          0  IR-PCI-MSI 499719-edge      enp0s30f4:tx-3
[ snip ]

$ ping -I enp0s30f4 192.168.1.10 -i 0.01 -c 100 > /dev/null

Statistic after ping:
---------------------
$ ethtool -S enp0s30f4

[ snip ]
     q3_tx_pkt_n: 8016
     q3_tx_irq_n: 320
[ snip ]

$ cat /proc/interrupts

[ snip ]
143:          0          0          0        320          0          0

         0          0  IR-PCI-MSI 499719-edge      enp0s30f4:tx-3
[ snip ]
====================
Signed-off-by: David S. Miller <davem@davemloft.net>

drivers/net/ethernet/stmicro/stmmac/common.h

@@ -58,6 +58,16 @@
 #undef FRAME_FILTER_DEBUG
 /* #define FRAME_FILTER_DEBUG */
 
+struct stmmac_txq_stats {
+	unsigned long tx_pkt_n;
+	unsigned long tx_normal_irq_n;
+};
+
+struct stmmac_rxq_stats {
+	unsigned long rx_pkt_n;
+	unsigned long rx_normal_irq_n;
+};
+
 /* Extra statistic and debug information exposed by ethtool */
 struct stmmac_extra_stats {
 	/* Transmit errors */
@@ -189,6 +199,9 @@ struct stmmac_extra_stats {
 	unsigned long mtl_est_hlbf;
 	unsigned long mtl_est_btre;
 	unsigned long mtl_est_btrlm;
+	/* per queue statistics */
+	struct stmmac_txq_stats txq_stats[MTL_MAX_TX_QUEUES];
+	struct stmmac_rxq_stats rxq_stats[MTL_MAX_RX_QUEUES];
 };
 
 /* Safety Feature statistics exposed by ethtool */

drivers/net/ethernet/stmicro/stmmac/dwmac4_lib.c

@@ -170,13 +170,16 @@ int dwmac4_dma_interrupt(void __iomem *ioaddr,
 		x->normal_irq_n++;
 	if (likely(intr_status & DMA_CHAN_STATUS_RI)) {
 		x->rx_normal_irq_n++;
+		x->rxq_stats[chan].rx_normal_irq_n++;
 		ret |= handle_rx;
 	}
-	if (likely(intr_status & (DMA_CHAN_STATUS_TI |
-		DMA_CHAN_STATUS_TBU))) {
+	if (likely(intr_status & DMA_CHAN_STATUS_TI)) {
 		x->tx_normal_irq_n++;
+		x->txq_stats[chan].tx_normal_irq_n++;
 		ret |= handle_tx;
 	}
+	if (unlikely(intr_status & DMA_CHAN_STATUS_TBU))
+		ret |= handle_tx;
 	if (unlikely(intr_status & DMA_CHAN_STATUS_ERI))
 		x->rx_early_irq++;
 

drivers/net/ethernet/stmicro/stmmac/stmmac_ethtool.c

@@ -261,6 +261,18 @@ static const struct stmmac_stats stmmac_mmc[] = {
 };
 #define STMMAC_MMC_STATS_LEN ARRAY_SIZE(stmmac_mmc)
 
+static const char stmmac_qstats_tx_string[][ETH_GSTRING_LEN] = {
+	"tx_pkt_n",
+	"tx_irq_n",
+#define STMMAC_TXQ_STATS ARRAY_SIZE(stmmac_qstats_tx_string)
+};
+
+static const char stmmac_qstats_rx_string[][ETH_GSTRING_LEN] = {
+	"rx_pkt_n",
+	"rx_irq_n",
+#define STMMAC_RXQ_STATS ARRAY_SIZE(stmmac_qstats_rx_string)
+};
+
 static void stmmac_ethtool_getdrvinfo(struct net_device *dev,
 				      struct ethtool_drvinfo *info)
 {
@@ -510,6 +522,31 @@ stmmac_set_pauseparam(struct net_device *netdev,
 	}
 }
 
+static void stmmac_get_per_qstats(struct stmmac_priv *priv, u64 *data)
+{
+	u32 tx_cnt = priv->plat->tx_queues_to_use;
+	u32 rx_cnt = priv->plat->rx_queues_to_use;
+	int q, stat;
+	char *p;
+
+	for (q = 0; q < tx_cnt; q++) {
+		p = (char *)priv + offsetof(struct stmmac_priv,
+					    xstats.txq_stats[q].tx_pkt_n);
+		for (stat = 0; stat < STMMAC_TXQ_STATS; stat++) {
+			*data++ = (*(u64 *)p);
+			p += sizeof(u64 *);
+		}
+	}
+	for (q = 0; q < rx_cnt; q++) {
+		p = (char *)priv + offsetof(struct stmmac_priv,
+					    xstats.rxq_stats[q].rx_pkt_n);
+		for (stat = 0; stat < STMMAC_RXQ_STATS; stat++) {
+			*data++ = (*(u64 *)p);
+			p += sizeof(u64 *);
+		}
+	}
+}
+
 static void stmmac_get_ethtool_stats(struct net_device *dev,
 				 struct ethtool_stats *dummy, u64 *data)
 {
@@ -560,16 +597,21 @@ static void stmmac_get_ethtool_stats(struct net_device *dev,
 		data[j++] = (stmmac_gstrings_stats[i].sizeof_stat ==
 			     sizeof(u64)) ? (*(u64 *)p) : (*(u32 *)p);
 	}
+	stmmac_get_per_qstats(priv, &data[j]);
 }
 
 static int stmmac_get_sset_count(struct net_device *netdev, int sset)
 {
 	struct stmmac_priv *priv = netdev_priv(netdev);
+	u32 tx_cnt = priv->plat->tx_queues_to_use;
+	u32 rx_cnt = priv->plat->rx_queues_to_use;
 	int i, len, safety_len = 0;
 
 	switch (sset) {
 	case ETH_SS_STATS:
-		len = STMMAC_STATS_LEN;
+		len = STMMAC_STATS_LEN +
+		      STMMAC_TXQ_STATS * tx_cnt +
+		      STMMAC_RXQ_STATS * rx_cnt;
 
 		if (priv->dma_cap.rmon)
 			len += STMMAC_MMC_STATS_LEN;
@@ -592,6 +634,28 @@ static int stmmac_get_sset_count(struct net_device *netdev, int sset)
 	}
 }
 
+static void stmmac_get_qstats_string(struct stmmac_priv *priv, u8 *data)
+{
+	u32 tx_cnt = priv->plat->tx_queues_to_use;
+	u32 rx_cnt = priv->plat->rx_queues_to_use;
+	int q, stat;
+
+	for (q = 0; q < tx_cnt; q++) {
+		for (stat = 0; stat < STMMAC_TXQ_STATS; stat++) {
+			snprintf(data, ETH_GSTRING_LEN, "q%d_%s", q,
+				 stmmac_qstats_tx_string[stat]);
+			data += ETH_GSTRING_LEN;
+		}
+	}
+	for (q = 0; q < rx_cnt; q++) {
+		for (stat = 0; stat < STMMAC_RXQ_STATS; stat++) {
+			snprintf(data, ETH_GSTRING_LEN, "q%d_%s", q,
+				 stmmac_qstats_rx_string[stat]);
+			data += ETH_GSTRING_LEN;
+		}
+	}
+}
+
 static void stmmac_get_strings(struct net_device *dev, u32 stringset, u8 *data)
 {
 	int i;
@@ -622,6 +686,7 @@ static void stmmac_get_strings(struct net_device *dev, u32 stringset, u8 *data)
 				ETH_GSTRING_LEN);
 			p += ETH_GSTRING_LEN;
 		}
+		stmmac_get_qstats_string(priv, p);
 		break;
 	case ETH_SS_TEST:
 		stmmac_selftest_get_strings(priv, p);

drivers/net/ethernet/stmicro/stmmac/stmmac_main.c

@@ -2500,6 +2500,7 @@ static int stmmac_tx_clean(struct stmmac_priv *priv, int budget, u32 queue)
 			} else {
 				priv->dev->stats.tx_packets++;
 				priv->xstats.tx_pkt_n++;
+				priv->xstats.txq_stats[queue].tx_pkt_n++;
 			}
 			if (skb)
 				stmmac_get_tx_hwtstamp(priv, p, skb);
@@ -5000,6 +5001,9 @@ static int stmmac_rx_zc(struct stmmac_priv *priv, int limit, u32 queue)
 
 	stmmac_finalize_xdp_rx(priv, xdp_status);
 
+	priv->xstats.rx_pkt_n += count;
+	priv->xstats.rxq_stats[queue].rx_pkt_n += count;
+
 	if (xsk_uses_need_wakeup(rx_q->xsk_pool)) {
 		if (failure || stmmac_rx_dirty(priv, queue) > 0)
 			xsk_set_rx_need_wakeup(rx_q->xsk_pool);
@@ -5287,6 +5291,7 @@ static int stmmac_rx(struct stmmac_priv *priv, int limit, u32 queue)
 	stmmac_rx_refill(priv, queue);
 
 	priv->xstats.rx_pkt_n += count;
+	priv->xstats.rxq_stats[queue].rx_pkt_n += count;
 
 	return count;
 }