path: root/packages/linux/linux-2.6.27+2.6.28-rc6/afeb9260/0002-SRAM-TX-buffers-implementation-from-atmel-to-fix-TX.patch

From 856f17b89eb0e943d36f36d0c760e1226db9d716 Mon Sep 17 00:00:00 2001
From: Sergey Lapin <slapin@ossfans.org>
Date: Fri, 4 Jul 2008 01:52:39 +0400
Subject: [PATCH] SRAM TX buffers implementation from atmel to fix TX underrun errors

Signed-off-by: Sergey Lapin <slapin@ossfans.org>
---
 arch/arm/mach-at91/at91sam9260_devices.c |    7 ++
 drivers/net/Kconfig                      |    6 ++
 drivers/net/macb.c                       |  108 ++++++++++++++++++++++--------
 drivers/net/macb.h                       |    7 ++-
 4 files changed, 99 insertions(+), 29 deletions(-)

diff --git a/arch/arm/mach-at91/at91sam9260_devices.c b/arch/arm/mach-at91/at91sam9260_devices.c
index f5fec0a..681cd40 100644
--- a/arch/arm/mach-at91/at91sam9260_devices.c
+++ b/arch/arm/mach-at91/at91sam9260_devices.c
@@ -141,6 +141,13 @@ static struct resource eth_resources[] = {
 		.end	= AT91SAM9260_ID_EMAC,
 		.flags	= IORESOURCE_IRQ,
 	},
+#if defined(CONFIG_MACB_TX_SRAM)
+	[2] = {
+		.start	= AT91SAM9260_SRAM1_BASE,
+		.end	= AT91SAM9260_SRAM1_BASE + AT91SAM9260_SRAM1_SIZE - 1,
+		.flags	= IORESOURCE_MEM,
+	},
+#endif
 };
 
 static struct platform_device at91sam9260_eth_device = {
diff --git a/drivers/net/Kconfig b/drivers/net/Kconfig
index 8a03875..a9a46a8 100644
--- a/drivers/net/Kconfig
+++ b/drivers/net/Kconfig
@@ -218,6 +218,12 @@ config MACB
 	  To compile this driver as a module, choose M here: the module
 	  will be called macb.
 
+config MACB_TX_SRAM
+	bool "Use internal SRAM for EMAC (Ethernet) transmit buffers"
+	depends on (ARCH_AT91SAM9263 || ARCH_AT91SAM9260) && MACB
+	help
+	  Select this if you are using internal SRAM for EMAC transmit buffers.
+
 source "drivers/net/arm/Kconfig"
 
 config AX88796
diff --git a/drivers/net/macb.c b/drivers/net/macb.c
index daba82b..55063dc 100644
--- a/drivers/net/macb.c
+++ b/drivers/net/macb.c
@@ -33,22 +33,27 @@
 /* Make the IP header word-aligned (the ethernet header is 14 bytes) */
 #define RX_OFFSET		2
 
-#define TX_RING_SIZE		128
-#define DEF_TX_RING_PENDING	(TX_RING_SIZE - 1)
-#define TX_RING_BYTES		(sizeof(struct dma_desc) * TX_RING_SIZE)
+#define TX_MAX_PKT_SIZE		1536
+#define TX_DEFAULT_RING_SIZE	128 
+#define DEF_TX_RING_PENDING(bp)		((bp)->tx_ring_size - 1)
+
+#define TX_RING_BYTES(bp)		(sizeof(struct dma_desc) * \
+						(bp)->tx_ring_size)
 
 #define TX_RING_GAP(bp)						\
-	(TX_RING_SIZE - (bp)->tx_pending)
+	((bp)->tx_ring_size - (bp)->tx_pending)
+
 #define TX_BUFFS_AVAIL(bp)					\
 	(((bp)->tx_tail <= (bp)->tx_head) ?			\
 	 (bp)->tx_tail + (bp)->tx_pending - (bp)->tx_head :	\
 	 (bp)->tx_tail - (bp)->tx_head - TX_RING_GAP(bp))
-#define NEXT_TX(n)		(((n) + 1) & (TX_RING_SIZE - 1))
 
-#define NEXT_RX(n)		(((n) + 1) & (RX_RING_SIZE - 1))
+#define NEXT_TX(bp, n)         (((n) + 1) % ((bp)->tx_ring_size)) 
+#define NEXT_RX(n)             (((n) + 1) % (RX_RING_SIZE)) 
+ 
 
 /* minimum number of free TX descriptors before waking up TX process */
-#define MACB_TX_WAKEUP_THRESH	(TX_RING_SIZE / 4)
+#define MACB_TX_WAKEUP_THRESH(bp)	((bp)->tx_ring_size / 4)
 
 #define MACB_RX_INT_FLAGS	(MACB_BIT(RCOMP) | MACB_BIT(RXUBR)	\
 				 | MACB_BIT(ISR_ROVR))
@@ -316,11 +321,11 @@ static void macb_tx(struct macb *bp)
 		head = bp->tx_head;
 
 		/*Mark all the buffer as used to avoid sending a lost buffer*/
-		for (i = 0; i < TX_RING_SIZE; i++)
+		for (i = 0; i < bp->tx_ring_size; i++)
 			bp->tx_ring[i].ctrl = MACB_BIT(TX_USED);
 
 		/* free transmit buffer in upper layer*/
-		for (tail = bp->tx_tail; tail != head; tail = NEXT_TX(tail)) {
+		for (tail = bp->tx_tail; tail != head; tail = NEXT_TX(bp, tail)) {
 			struct ring_info *rp = &bp->tx_skb[tail];
 			struct sk_buff *skb = rp->skb;
 
@@ -346,7 +351,7 @@ static void macb_tx(struct macb *bp)
 		return;
 
 	head = bp->tx_head;
-	for (tail = bp->tx_tail; tail != head; tail = NEXT_TX(tail)) {
+	for (tail = bp->tx_tail; tail != head; tail = NEXT_TX(bp, tail)) {
 		struct ring_info *rp = &bp->tx_skb[tail];
 		struct sk_buff *skb = rp->skb;
 		u32 bufstat;
@@ -371,7 +376,7 @@ static void macb_tx(struct macb *bp)
 
 	bp->tx_tail = tail;
 	if (netif_queue_stopped(bp->dev) &&
-	    TX_BUFFS_AVAIL(bp) > MACB_TX_WAKEUP_THRESH)
+	    TX_BUFFS_AVAIL(bp) > MACB_TX_WAKEUP_THRESH(bp))
 		netif_wake_queue(bp->dev);
 }
 
@@ -401,6 +406,7 @@ static int macb_rx_frame(struct macb *bp, unsigned int first_frag,
 	}
 
 	skb_reserve(skb, RX_OFFSET);
+	skb->dev = bp->dev;
 	skb->ip_summed = CHECKSUM_NONE;
 	skb_put(skb, len);
 
@@ -637,8 +643,15 @@ static int macb_start_xmit(struct sk_buff *skb, struct net_device *dev)
 
 	entry = bp->tx_head;
 	dev_dbg(&bp->pdev->dev, "Allocated ring entry %u\n", entry);
-	mapping = dma_map_single(&bp->pdev->dev, skb->data,
+	if (bp->smem_start) {
+		mapping = bp->tx_ring[entry].addr;
+		memcpy(bp->tx_buffers + entry * TX_MAX_PKT_SIZE,
+			skb->data, len);
+	} else {
+		mapping = dma_map_single(&bp->pdev->dev, skb->data,
 				 len, DMA_TO_DEVICE);
+		bp->tx_ring[entry].addr = mapping;
+	}
 	bp->tx_skb[entry].skb = skb;
 	bp->tx_skb[entry].mapping = mapping;
 	dev_dbg(&bp->pdev->dev, "Mapped skb data %p to DMA addr %08lx\n",
@@ -646,14 +659,13 @@ static int macb_start_xmit(struct sk_buff *skb, struct net_device *dev)
 
 	ctrl = MACB_BF(TX_FRMLEN, len);
 	ctrl |= MACB_BIT(TX_LAST);
-	if (entry == (TX_RING_SIZE - 1))
+	if (entry == (bp->tx_ring_size - 1))
 		ctrl |= MACB_BIT(TX_WRAP);
 
-	bp->tx_ring[entry].addr = mapping;
 	bp->tx_ring[entry].ctrl = ctrl;
 	wmb();
 
-	entry = NEXT_TX(entry);
+	entry = NEXT_TX(bp, entry);
 	bp->tx_head = entry;
 
 	macb_writel(bp, NCR, macb_readl(bp, NCR) | MACB_BIT(TSTART));
@@ -680,7 +692,8 @@ static void macb_free_consistent(struct macb *bp)
 		bp->rx_ring = NULL;
 	}
 	if (bp->tx_ring) {
-		dma_free_coherent(&bp->pdev->dev, TX_RING_BYTES,
+		if (!bp->smem_start)
+			dma_free_coherent(&bp->pdev->dev, TX_RING_BYTES(bp),
 				  bp->tx_ring, bp->tx_ring_dma);
 		bp->tx_ring = NULL;
 	}
@@ -690,13 +703,16 @@ static void macb_free_consistent(struct macb *bp)
 				  bp->rx_buffers, bp->rx_buffers_dma);
 		bp->rx_buffers = NULL;
 	}
+
+	if (bp->smem_start && bp->tx_ring_dma)
+		release_mem_region(bp->tx_ring_dma, bp->smem_size);
 }
 
 static int macb_alloc_consistent(struct macb *bp)
 {
 	int size;
 
-	size = TX_RING_SIZE * sizeof(struct ring_info);
+	size = bp->tx_ring_size * sizeof(struct ring_info);
 	bp->tx_skb = kmalloc(size, GFP_KERNEL);
 	if (!bp->tx_skb)
 		goto out_err;
@@ -710,14 +726,30 @@ static int macb_alloc_consistent(struct macb *bp)
 		"Allocated RX ring of %d bytes at %08lx (mapped %p)\n",
 		size, (unsigned long)bp->rx_ring_dma, bp->rx_ring);
 
-	size = TX_RING_BYTES;
-	bp->tx_ring = dma_alloc_coherent(&bp->pdev->dev, size,
+	if (bp->smem_start) {
+		if (request_mem_region(bp->smem_start,
+			bp->smem_size, "macb")) {
+			bp->tx_ring_dma = bp->smem_start;
+		}
+		bp->tx_ring = ioremap(bp->tx_ring_dma, bp->smem_size);
+		if (bp->tx_ring) {
+			bp->tx_buffers_dma = bp->tx_ring_dma +
+			TX_RING_BYTES(bp);
+			bp->tx_buffers = (char *) bp->tx_ring +
+			TX_RING_BYTES(bp);
+		}
+	}
+
+	if (!bp->tx_ring) {
+		size = TX_RING_BYTES(bp);
+		bp->tx_ring = dma_alloc_coherent(&bp->pdev->dev, size,
 					 &bp->tx_ring_dma, GFP_KERNEL);
-	if (!bp->tx_ring)
-		goto out_err;
-	dev_dbg(&bp->pdev->dev,
-		"Allocated TX ring of %d bytes at %08lx (mapped %p)\n",
-		size, (unsigned long)bp->tx_ring_dma, bp->tx_ring);
+		if (!bp->tx_ring)
+			goto out_err;
+		dev_dbg(&bp->pdev->dev,
+			"Allocated TX ring of %d bytes at %08lx (mapped %p)\n",
+			size, (unsigned long)bp->tx_ring_dma, bp->tx_ring);
+	}
 
 	size = RX_RING_SIZE * RX_BUFFER_SIZE;
 	bp->rx_buffers = dma_alloc_coherent(&bp->pdev->dev, size,
@@ -748,11 +780,14 @@ static void macb_init_rings(struct macb *bp)
 	}
 	bp->rx_ring[RX_RING_SIZE - 1].addr |= MACB_BIT(RX_WRAP);
 
-	for (i = 0; i < TX_RING_SIZE; i++) {
-		bp->tx_ring[i].addr = 0;
+	for (i = 0; i < bp->tx_ring_size; i++) {
+		bp->tx_ring[i].addr = (bp->smem_start) ?
+			bp->tx_buffers_dma + i * TX_MAX_PKT_SIZE : 0;
 		bp->tx_ring[i].ctrl = MACB_BIT(TX_USED);
+		bp->tx_ring[i].addr = (u32) (bp->tx_buffers_dma +
+			(i * TX_MAX_PKT_SIZE));
 	}
-	bp->tx_ring[TX_RING_SIZE - 1].ctrl |= MACB_BIT(TX_WRAP);
+	bp->tx_ring[bp->tx_ring_size - 1].ctrl |= MACB_BIT(TX_WRAP);
 
 	bp->rx_tail = bp->tx_head = bp->tx_tail = 0;
 }
@@ -1090,6 +1125,7 @@ static int __init macb_probe(struct platform_device *pdev)
 {
 	struct eth_platform_data *pdata;
 	struct resource *regs;
+	struct resource *smem;
 	struct net_device *dev;
 	struct macb *bp;
 	struct phy_device *phydev;
@@ -1201,7 +1237,23 @@ static int __init macb_probe(struct platform_device *pdev)
 		macb_writel(bp, USRIO, MACB_BIT(MII));
 #endif
 
-	bp->tx_pending = DEF_TX_RING_PENDING;
+	/* Check  mem region for TX buffers */
+	smem = platform_get_resource(pdev, IORESOURCE_MEM, 1);
+	if (smem) {
+		bp->smem_start = smem->start;
+		bp->smem_size  = smem->end - smem->start + 1;
+		bp->tx_ring_size = bp->smem_size / (TX_MAX_PKT_SIZE +
+				sizeof(struct dma_desc));
+	}
+
+	/* if ring_size == 0 then mem is too
+	  small and use standard memory */
+	if (!bp->tx_ring_size) {
+		bp->smem_start = 0;
+		bp->tx_ring_size = TX_DEFAULT_RING_SIZE;
+	}
+
+	bp->tx_pending = DEF_TX_RING_PENDING(bp);
 
 	err = register_netdev(dev);
 	if (err) {
diff --git a/drivers/net/macb.h b/drivers/net/macb.h
index 57b85ac..faa5a9d 100644
--- a/drivers/net/macb.h
+++ b/drivers/net/macb.h
@@ -364,11 +364,15 @@ struct macb {
 	unsigned int		rx_tail;
 	struct dma_desc		*rx_ring;
 	void			*rx_buffers;
-
+	void			*tx_buffers;
+	unsigned int		tx_ring_size;
 	unsigned int		tx_head, tx_tail;
 	struct dma_desc		*tx_ring;
 	struct ring_info	*tx_skb;
 
+	unsigned int		smem_start;
+	unsigned int		smem_size;
+
 	spinlock_t		lock;
 	struct platform_device	*pdev;
 	struct clk		*pclk;
@@ -381,6 +385,7 @@ struct macb {
 	dma_addr_t		rx_ring_dma;
 	dma_addr_t		tx_ring_dma;
 	dma_addr_t		rx_buffers_dma;
+	dma_addr_t		tx_buffers_dma;
 
 	unsigned int		rx_pending, tx_pending;
 
-- 
1.5.4.1