Merge branch 'gve-improvements'

Jeroen de Borst says:

====================
gve: minor code and performance improvements

This patchset contains a number of independent minor code and performance
improvements.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>

Author: davem330

drivers/net/ethernet/google/gve/gve.h

@@ -30,7 +30,7 @@
 #define GVE_MIN_MSIX 3
 
 /* Numbers of gve tx/rx stats in stats report. */
-#define GVE_TX_STATS_REPORT_NUM	5
+#define GVE_TX_STATS_REPORT_NUM	6
 #define GVE_RX_STATS_REPORT_NUM	2
 
 /* Interval to schedule a stats report update, 20000ms. */
@@ -341,8 +341,8 @@ struct gve_tx_ring {
 	union {
 		/* GQI fields */
 		struct {
-			/* NIC tail pointer */
-			__be32 last_nic_done;
+			/* Spinlock for when cleanup in progress */
+			spinlock_t clean_lock;
 		};
 
 		/* DQO fields. */
@@ -413,7 +413,9 @@ struct gve_tx_ring {
 	u32 q_num ____cacheline_aligned; /* queue idx */
 	u32 stop_queue; /* count of queue stops */
 	u32 wake_queue; /* count of queue wakes */
+	u32 queue_timeout; /* count of queue timeouts */
 	u32 ntfy_id; /* notification block index */
+	u32 last_kick_msec; /* Last time the queue was kicked */
 	dma_addr_t bus; /* dma address of the descr ring */
 	dma_addr_t q_resources_bus; /* dma address of the queue resources */
 	dma_addr_t complq_bus_dqo; /* dma address of the dqo.compl_ring */
@@ -821,15 +823,15 @@ netdev_tx_t gve_tx(struct sk_buff *skb, struct net_device *dev);
 bool gve_tx_poll(struct gve_notify_block *block, int budget);
 int gve_tx_alloc_rings(struct gve_priv *priv);
 void gve_tx_free_rings_gqi(struct gve_priv *priv);
-__be32 gve_tx_load_event_counter(struct gve_priv *priv,
-				 struct gve_tx_ring *tx);
+u32 gve_tx_load_event_counter(struct gve_priv *priv,
+			      struct gve_tx_ring *tx);
+bool gve_tx_clean_pending(struct gve_priv *priv, struct gve_tx_ring *tx);
 /* rx handling */
 void gve_rx_write_doorbell(struct gve_priv *priv, struct gve_rx_ring *rx);
-bool gve_rx_poll(struct gve_notify_block *block, int budget);
+int gve_rx_poll(struct gve_notify_block *block, int budget);
+bool gve_rx_work_pending(struct gve_rx_ring *rx);
 int gve_rx_alloc_rings(struct gve_priv *priv);
 void gve_rx_free_rings_gqi(struct gve_priv *priv);
-bool gve_clean_rx_done(struct gve_rx_ring *rx, int budget,
-		       netdev_features_t feat);
 /* Reset */
 void gve_schedule_reset(struct gve_priv *priv);
 int gve_reset(struct gve_priv *priv, bool attempt_teardown);

drivers/net/ethernet/google/gve/gve_adminq.h

@@ -270,6 +270,7 @@ enum gve_stat_names {
 	TX_LAST_COMPLETION_PROCESSED	= 5,
 	RX_NEXT_EXPECTED_SEQUENCE	= 6,
 	RX_BUFFERS_POSTED		= 7,
+	TX_TIMEOUT_CNT			= 8,
 	// stats from NIC
 	RX_QUEUE_DROP_CNT		= 65,
 	RX_NO_BUFFERS_POSTED		= 66,

drivers/net/ethernet/google/gve/gve_ethtool.c

@@ -330,8 +330,7 @@ gve_get_ethtool_stats(struct net_device *netdev,
 			data[i++] = tmp_tx_bytes;
 			data[i++] = tx->wake_queue;
 			data[i++] = tx->stop_queue;
-			data[i++] = be32_to_cpu(gve_tx_load_event_counter(priv,
-									  tx));
+			data[i++] = gve_tx_load_event_counter(priv, tx);
 			data[i++] = tx->dma_mapping_error;
 			/* stats from NIC */
 			if (skip_nic_stats) {

drivers/net/ethernet/google/gve/gve_main.c

@@ -24,6 +24,9 @@
 #define GVE_VERSION		"1.0.0"
 #define GVE_VERSION_PREFIX	"GVE-"
 
+// Minimum amount of time between queue kicks in msec (10 seconds)
+#define MIN_TX_TIMEOUT_GAP (1000 * 10)
+
 const char gve_version_str[] = GVE_VERSION;
 static const char gve_version_prefix[] = GVE_VERSION_PREFIX;
 
@@ -192,34 +195,40 @@ static int gve_napi_poll(struct napi_struct *napi, int budget)
 	__be32 __iomem *irq_doorbell;
 	bool reschedule = false;
 	struct gve_priv *priv;
+	int work_done = 0;
 
 	block = container_of(napi, struct gve_notify_block, napi);
 	priv = block->priv;
 
 	if (block->tx)
 		reschedule |= gve_tx_poll(block, budget);
-	if (block->rx)
-		reschedule |= gve_rx_poll(block, budget);
+	if (block->rx) {
+		work_done = gve_rx_poll(block, budget);
+		reschedule |= work_done == budget;
+	}
 
 	if (reschedule)
 		return budget;
 
-	napi_complete(napi);
-	irq_doorbell = gve_irq_doorbell(priv, block);
-	iowrite32be(GVE_IRQ_ACK | GVE_IRQ_EVENT, irq_doorbell);
+       /* Complete processing - don't unmask irq if busy polling is enabled */
+	if (likely(napi_complete_done(napi, work_done))) {
+		irq_doorbell = gve_irq_doorbell(priv, block);
+		iowrite32be(GVE_IRQ_ACK | GVE_IRQ_EVENT, irq_doorbell);
 
-	/* Double check we have no extra work.
-	 * Ensure unmask synchronizes with checking for work.
-	 */
-	mb();
-	if (block->tx)
-		reschedule |= gve_tx_poll(block, -1);
-	if (block->rx)
-		reschedule |= gve_rx_poll(block, -1);
-	if (reschedule && napi_reschedule(napi))
-		iowrite32be(GVE_IRQ_MASK, irq_doorbell);
+		/* Ensure IRQ ACK is visible before we check pending work.
+		 * If queue had issued updates, it would be truly visible.
+		 */
+		mb();
 
-	return 0;
+		if (block->tx)
+			reschedule |= gve_tx_clean_pending(priv, block->tx);
+		if (block->rx)
+			reschedule |= gve_rx_work_pending(block->rx);
+
+		if (reschedule && napi_reschedule(napi))
+			iowrite32be(GVE_IRQ_MASK, irq_doorbell);
+	}
+	return work_done;
 }
 
 static int gve_napi_poll_dqo(struct napi_struct *napi, int budget)
@@ -1115,9 +1124,47 @@ static void gve_turnup(struct gve_priv *priv)
 
 static void gve_tx_timeout(struct net_device *dev, unsigned int txqueue)
 {
-	struct gve_priv *priv = netdev_priv(dev);
+	struct gve_notify_block *block;
+	struct gve_tx_ring *tx = NULL;
+	struct gve_priv *priv;
+	u32 last_nic_done;
+	u32 current_time;
+	u32 ntfy_idx;
 
+	netdev_info(dev, "Timeout on tx queue, %d", txqueue);
+	priv = netdev_priv(dev);
+	if (txqueue > priv->tx_cfg.num_queues)
+		goto reset;
+
+	ntfy_idx = gve_tx_idx_to_ntfy(priv, txqueue);
+	if (ntfy_idx > priv->num_ntfy_blks)
+		goto reset;
+
+	block = &priv->ntfy_blocks[ntfy_idx];
+	tx = block->tx;
+
+	current_time = jiffies_to_msecs(jiffies);
+	if (tx->last_kick_msec + MIN_TX_TIMEOUT_GAP > current_time)
+		goto reset;
+
+	/* Check to see if there are missed completions, which will allow us to
+	 * kick the queue.
+	 */
+	last_nic_done = gve_tx_load_event_counter(priv, tx);
+	if (last_nic_done - tx->done) {
+		netdev_info(dev, "Kicking queue %d", txqueue);
+		iowrite32be(GVE_IRQ_MASK, gve_irq_doorbell(priv, block));
+		napi_schedule(&block->napi);
+		tx->last_kick_msec = current_time;
+		goto out;
+	} // Else reset.
+
+reset:
 	gve_schedule_reset(priv);
+
+out:
+	if (tx)
+		tx->queue_timeout++;
 	priv->tx_timeo_cnt++;
 }
 
@@ -1246,6 +1293,11 @@ void gve_handle_report_stats(struct gve_priv *priv)
 				.value = cpu_to_be64(last_completion),
 				.queue_id = cpu_to_be32(idx),
 			};
+			stats[stats_idx++] = (struct stats) {
+				.stat_name = cpu_to_be32(TX_TIMEOUT_CNT),
+				.value = cpu_to_be64(priv->tx[idx].queue_timeout),
+				.queue_id = cpu_to_be32(idx),
+			};
 		}
 	}
 	/* rx stats */