net/macb: better manage tx errors

Handle all TX errors, not only underruns. TX error management is
deferred to a dedicated workqueue.
Reinitialize the TX ring after treating all remaining frames, and
restart the controller when everything has been cleaned up properly.
Napi is not stopped during this task as the driver only handles
napi for RX for now.
With this sequence, we do not need a special check during the xmit
method as the packets will be caught by TX disable during workqueue
execution.
Signed-off-by: Nicolas Ferre <nicolas.ferre@atmel.com>
Tested-by: Joachim Eastwood <manabian@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

drivers/net/ethernet/cadence/macb.c

@@ -44,6 +44,16 @@
 
 #define MACB_RX_INT_FLAGS	(MACB_BIT(RCOMP) | MACB_BIT(RXUBR)	\
 				 | MACB_BIT(ISR_ROVR))
+#define MACB_TX_ERR_FLAGS	(MACB_BIT(ISR_TUND)			\
+					| MACB_BIT(ISR_RLE)		\
+					| MACB_BIT(TXERR))
+#define MACB_TX_INT_FLAGS	(MACB_TX_ERR_FLAGS | MACB_BIT(TCOMP))
+
+/*
+ * Graceful stop timeouts in us. We should allow up to
+ * 1 frame time (10 Mbits/s, full-duplex, ignoring collisions)
+ */
+#define MACB_HALT_TIMEOUT	1230
 
 /* Ring buffer accessors */
 static unsigned int macb_tx_ring_wrap(unsigned int index)
@@ -339,66 +349,113 @@ static void macb_update_stats(struct macb *bp)
 		*p += __raw_readl(reg);
 }
 
-static void macb_tx(struct macb *bp)
+static int macb_halt_tx(struct macb *bp)
 {
-	unsigned int tail;
-	unsigned int head;
+	unsigned long	halt_time, timeout;
 	u32		status;
 
+	macb_writel(bp, NCR, macb_readl(bp, NCR) | MACB_BIT(THALT));
+
+	timeout = jiffies + usecs_to_jiffies(MACB_HALT_TIMEOUT);
+	do {
+		halt_time = jiffies;
 		status = macb_readl(bp, TSR);
-	macb_writel(bp, TSR, status);
+		if (!(status & MACB_BIT(TGO)))
+			return 0;
 
-	netdev_vdbg(bp->dev, "macb_tx status = 0x%03lx\n", (unsigned long)status);
+		usleep_range(10, 250);
+	} while (time_before(halt_time, timeout));
 
-	if (status & (MACB_BIT(UND) | MACB_BIT(TSR_RLE))) {
-		int i;
-		netdev_err(bp->dev, "TX %s, resetting buffers\n",
-			   status & MACB_BIT(UND) ?
-			   "underrun" : "retry limit exceeded");
+	return -ETIMEDOUT;
+}
 
-		/* Transfer ongoing, disable transmitter, to avoid confusion */
-		if (status & MACB_BIT(TGO))
-			macb_writel(bp, NCR, macb_readl(bp, NCR) & ~MACB_BIT(TE));
+static void macb_tx_error_task(struct work_struct *work)
+{
+	struct macb	*bp = container_of(work, struct macb, tx_error_task);
+	struct macb_tx_skb	*tx_skb;
+	struct sk_buff		*skb;
+	unsigned int		tail;
 
-		head = bp->tx_head;
+	netdev_vdbg(bp->dev, "macb_tx_error_task: t = %u, h = %u\n",
+		    bp->tx_tail, bp->tx_head);
 
-		/*Mark all the buffer as used to avoid sending a lost buffer*/
-		for (i = 0; i < TX_RING_SIZE; i++)
-			bp->tx_ring[i].ctrl = MACB_BIT(TX_USED);
+	/* Make sure nobody is trying to queue up new packets */
+	netif_stop_queue(bp->dev);
 
-		/* Add wrap bit */
-		bp->tx_ring[TX_RING_SIZE - 1].ctrl |= MACB_BIT(TX_WRAP);
+	/*
+	 * Stop transmission now
+	 * (in case we have just queued new packets)
+	 */
+	if (macb_halt_tx(bp))
+		/* Just complain for now, reinitializing TX path can be good */
+		netdev_err(bp->dev, "BUG: halt tx timed out\n");
 
-		/* free transmit buffer in upper layer*/
-		for (tail = bp->tx_tail; tail != head; tail++) {
-			struct macb_tx_skb	*tx_skb;
-			struct sk_buff		*skb;
+	/* No need for the lock here as nobody will interrupt us anymore */
 
-			rmb();
+	/*
+	 * Treat frames in TX queue including the ones that caused the error.
+	 * Free transmit buffers in upper layer.
+	 */
+	for (tail = bp->tx_tail; tail != bp->tx_head; tail++) {
+		struct macb_dma_desc	*desc;
+		u32			ctrl;
 
+		desc = macb_tx_desc(bp, tail);
+		ctrl = desc->ctrl;
 		tx_skb = macb_tx_skb(bp, tail);
 		skb = tx_skb->skb;
 
-			dma_unmap_single(&bp->pdev->dev, tx_skb->mapping,
-						skb->len, DMA_TO_DEVICE);
+		if (ctrl & MACB_BIT(TX_USED)) {
+			netdev_vdbg(bp->dev, "txerr skb %u (data %p) TX complete\n",
+				    macb_tx_ring_wrap(tail), skb->data);
+			bp->stats.tx_packets++;
+			bp->stats.tx_bytes += skb->len;
+		} else {
+			/*
+			 * "Buffers exhausted mid-frame" errors may only happen
+			 * if the driver is buggy, so complain loudly about those.
+			 * Statistics are updated by hardware.
+			 */
+			if (ctrl & MACB_BIT(TX_BUF_EXHAUSTED))
+				netdev_err(bp->dev,
+					   "BUG: TX buffers exhausted mid-frame\n");
+
+			desc->ctrl = ctrl | MACB_BIT(TX_USED);
+		}
+
+		dma_unmap_single(&bp->pdev->dev, tx_skb->mapping, skb->len,
+				 DMA_TO_DEVICE);
 		tx_skb->skb = NULL;
-			dev_kfree_skb_irq(skb);
+		dev_kfree_skb(skb);
 	}
 
+	/* Make descriptor updates visible to hardware */
+	wmb();
+
+	/* Reinitialize the TX desc queue */
+	macb_writel(bp, TBQP, bp->tx_ring_dma);
+	/* Make TX ring reflect state of hardware */
 	bp->tx_head = bp->tx_tail = 0;
 
-		/* Enable the transmitter again */
-		if (status & MACB_BIT(TGO))
-			macb_writel(bp, NCR, macb_readl(bp, NCR) | MACB_BIT(TE));
-	}
+	/* Now we are ready to start transmission again */
+	netif_wake_queue(bp->dev);
 
-	if (!(status & MACB_BIT(COMP)))
-		/*
-		 * This may happen when a buffer becomes complete
-		 * between reading the ISR and scanning the
-		 * descriptors.  Nothing to worry about.
-		 */
-		return;
+	/* Housework before enabling TX IRQ */
+	macb_writel(bp, TSR, macb_readl(bp, TSR));
+	macb_writel(bp, IER, MACB_TX_INT_FLAGS);
+}
+
+static void macb_tx_interrupt(struct macb *bp)
+{
+	unsigned int tail;
+	unsigned int head;
+	u32 status;
+
+	status = macb_readl(bp, TSR);
+	macb_writel(bp, TSR, status);
+
+	netdev_vdbg(bp->dev, "macb_tx_interrupt status = 0x%03lx\n",
+		(unsigned long)status);
 
 	head = bp->tx_head;
 	for (tail = bp->tx_tail; tail != head; tail++) {
@@ -638,9 +695,14 @@ static irqreturn_t macb_interrupt(int irq, void *dev_id)
 			}
 		}
 
-		if (status & (MACB_BIT(TCOMP) | MACB_BIT(ISR_TUND) |
-			    MACB_BIT(ISR_RLE)))
-			macb_tx(bp);
+		if (unlikely(status & (MACB_TX_ERR_FLAGS))) {
+			macb_writel(bp, IDR, MACB_TX_INT_FLAGS);
+			schedule_work(&bp->tx_error_task);
+			break;
+		}
+
+		if (status & MACB_BIT(TCOMP))
+			macb_tx_interrupt(bp);
 
 		/*
 		 * Link change detection isn't possible with RMII, so we'll
@@ -970,13 +1032,8 @@ static void macb_init_hw(struct macb *bp)
 	macb_writel(bp, NCR, MACB_BIT(RE) | MACB_BIT(TE) | MACB_BIT(MPE));
 
 	/* Enable interrupts */
-	macb_writel(bp, IER, (MACB_BIT(RCOMP)
-			      | MACB_BIT(RXUBR)
-			      | MACB_BIT(ISR_TUND)
-			      | MACB_BIT(ISR_RLE)
-			      | MACB_BIT(TXERR)
-			      | MACB_BIT(TCOMP)
-			      | MACB_BIT(ISR_ROVR)
+	macb_writel(bp, IER, (MACB_RX_INT_FLAGS
+			      | MACB_TX_INT_FLAGS
 			      | MACB_BIT(HRESP)));
 
 }
@@ -1428,6 +1485,7 @@ static int __init macb_probe(struct platform_device *pdev)
 	bp->dev = dev;
 
 	spin_lock_init(&bp->lock);
+	INIT_WORK(&bp->tx_error_task, macb_tx_error_task);
 
 	bp->pclk = clk_get(&pdev->dev, "pclk");
 	if (IS_ERR(bp->pclk)) {

drivers/net/ethernet/cadence/macb.h

@@ -538,6 +538,7 @@ struct macb {
 	struct clk		*hclk;
 	struct net_device	*dev;
 	struct napi_struct	napi;
+	struct work_struct	tx_error_task;
 	struct net_device_stats	stats;
 	union {
 		struct macb_stats	macb;