atmel SoC SPI controller with dmaengine and DT

index b80891b..31ed405 100644
@@ -130,6 +130,21 @@ void of_device_make_bus_id(struct device *dev)
 	dev_set_name(dev, "%s.%d", node->name, magic - 1);
 }
 
+static void of_get_dma_mask(struct device *dev, struct device_node *np)
+{
+	const __be32 *prop;
+	int len;
+
+	prop = of_get_property(np, "dma-mask", &len);
+
+	dev->dma_mask = &dev->coherent_dma_mask;
+
+	if (!prop)
+		dev->coherent_dma_mask = DMA_BIT_MASK(32);
+	else
+		dev->coherent_dma_mask = of_read_number(prop, len / 4);
+}
+
 /**
  * of_device_alloc - Allocate and initialize an of_device
  * @np: device node to assign to device

@@ -171,10 +186,8 @@ struct platform_device *of_device_alloc(struct device_node *np,
 		WARN_ON(of_irq_to_resource_table(np, res, num_irq) != num_irq);
 	}
 
+	of_get_dma_mask(&dev->dev, np);
 	dev->dev.of_node = of_node_get(np);
-#if defined(CONFIG_MICROBLAZE)
-	dev->dev.dma_mask = &dev->archdata.dma_mask;
-#endif
 	dev->dev.parent = parent;
 
 	if (bus_id)

@@ -211,10 +224,6 @@ struct platform_device *of_platform_device_create_pdata(
 	if (!dev)
 		return NULL;
 
-#if defined(CONFIG_MICROBLAZE)
-	dev->archdata.dma_mask = 0xffffffffUL;
-#endif
-	dev->dev.coherent_dma_mask = DMA_BIT_MASK(32);
 	dev->dev.bus = &platform_bus_type;
 	dev->dev.platform_data = platform_data;
 

97 lines not on this page · raw message

index fa702ae..f629189 100644
@@ -90,6 +90,7 @@ struct spi_device {
 	void			*controller_state;
 	void			*controller_data;
 	char			modalias[SPI_NAME_SIZE];
+	int			cs_gpio;	/* chip select gpio */
 
 	/*
 	 * likely need more hooks for more protocol options affecting how

@@ -362,6 +363,8 @@ struct spi_master {
 	int (*transfer_one_message)(struct spi_master *master,
 				    struct spi_message *mesg);
 	int (*unprepare_transfer_hardware)(struct spi_master *master);
+	/* gpio chip select */
+	int			*cs_gpios;
 };
 
 static inline void *spi_master_get_devdata(struct spi_master *master)

index 16d6a83..2e040be 100644
@@ -188,6 +188,7 @@
 struct atmel_spi {
 	spinlock_t		lock;
 
+	resource_size_t		phybase;
 	void __iomem		*regs;
 	int			irq;
 	struct clk		*clk;

@@ -961,6 +962,7 @@ static int __devinit atmel_spi_probe(struct platform_device *pdev)
 	as->regs = ioremap(regs->start, resource_size(regs));
 	if (!as->regs)
 		goto out_free_buffer;
+	as->phybase = regs->start;
 	as->irq = irq;
 	as->clk = clk;
 

index 2e040be..fc1222a 100644
@@ -1010,6 +1010,7 @@ static int __devexit atmel_spi_remove(struct platform_device *pdev)
 	struct spi_master	*master = platform_get_drvdata(pdev);
 	struct atmel_spi	*as = spi_master_get_devdata(master);
 	struct spi_message	*msg;
+	struct spi_transfer	*xfer;
 
 	/* reset the hardware and block queue progress */
 	spin_lock_irq(&as->lock);

@@ -1021,9 +1022,10 @@ static int __devexit atmel_spi_remove(struct platform_device *pdev)
 
 	/* Terminate remaining queued transfers */
 	list_for_each_entry(msg, &as->queue, queue) {
-		/* REVISIT unmapping the dma is a NOP on ARM and AVR32
-		 * but we shouldn't depend on that...
-		 */
+		list_for_each_entry(xfer, &msg->transfers, transfer_list) {
+			if (!msg->is_dma_mapped)
+				atmel_spi_dma_unmap_xfer(master, xfer);
+		}
 		msg->status = -ESHUTDOWN;
 		msg->complete(msg->context);
 	}

index fc1222a..76a1baf 100644
@@ -201,6 +201,7 @@ struct atmel_spi {
 	unsigned long		current_remaining_bytes;
 	struct spi_transfer	*next_transfer;
 	unsigned long		next_remaining_bytes;
+	int			done_status;
 
 	void			*buffer;
 	dma_addr_t		buffer_dma;

@@ -544,15 +545,15 @@ static void atmel_spi_dma_unmap_xfer(struct spi_master *master,
 
 static void
 atmel_spi_msg_done(struct spi_master *master, struct atmel_spi *as,
-		struct spi_message *msg, int status, int stay)
+		struct spi_message *msg, int stay)
 {
-	if (!stay || status < 0)
+	if (!stay || as->done_status < 0)
 		cs_deactivate(as, msg->spi);
 	else
 		as->stay = msg->spi;
 
 	list_del(&msg->queue);
-	msg->status = status;
+	msg->status = as->done_status;
 
 	dev_dbg(master->dev.parent,
 		"xfer complete: %u bytes transferred\n",

@@ -564,6 +565,7 @@ atmel_spi_msg_done(struct spi_master *master, struct atmel_spi *as,
 
 	as->current_transfer = NULL;
 	as->next_transfer = NULL;
+	as->done_status = 0;
 
 	/* continue if needed */
 	if (list_empty(&as->queue) || as->stopping)

@@ -641,7 +643,8 @@ atmel_spi_interrupt(int irq, void *dev_id)
 		/* Clear any overrun happening while cleaning up */
 		spi_readl(as, SR);
 
-		atmel_spi_msg_done(master, as, msg, -EIO, 0);
+		as->done_status = -EIO;
+		atmel_spi_msg_done(master, as, msg, 0);
 	} else if (pending & (SPI_BIT(RXBUFF) | SPI_BIT(ENDRX))) {
 		ret = IRQ_HANDLED;
 

@@ -659,7 +662,7 @@ atmel_spi_interrupt(int irq, void *dev_id)
 
 			if (atmel_spi_xfer_is_last(msg, xfer)) {
 				/* report completed message */
-				atmel_spi_msg_done(master, as, msg, 0,
+				atmel_spi_msg_done(master, as, msg,
 						xfer->cs_change);
 			} else {
 				if (xfer->cs_change) {

index 76a1baf..37f54c3 100644
@@ -187,6 +187,7 @@
  */
 struct atmel_spi {
 	spinlock_t		lock;
+	unsigned long		flags;
 
 	resource_size_t		phybase;
 	void __iomem		*regs;

@@ -323,6 +324,16 @@ static void cs_deactivate(struct atmel_spi *as, struct spi_device *spi)
 		gpio_set_value(asd->npcs_pin, !active);
 }
 
+static void atmel_spi_lock(struct atmel_spi *as)
+{
+		spin_lock_irqsave(&as->lock, as->flags);
+}
+
+static void atmel_spi_unlock(struct atmel_spi *as)
+{
+		spin_unlock_irqrestore(&as->lock, as->flags);
+}
+
 static inline int atmel_spi_xfer_is_last(struct spi_message *msg,
 					struct spi_transfer *xfer)
 {

@@ -559,9 +570,9 @@ atmel_spi_msg_done(struct spi_master *master, struct atmel_spi *as,
 		"xfer complete: %u bytes transferred\n",
 		msg->actual_length);
 
-	spin_unlock(&as->lock);
+	atmel_spi_unlock(as);
 	msg->complete(msg->context);
-	spin_lock(&as->lock);
+	atmel_spi_lock(as);
 
 	as->current_transfer = NULL;
 	as->next_transfer = NULL;

@@ -788,13 +799,11 @@ static int atmel_spi_setup(struct spi_device *spi)
 		spi->controller_state = asd;
 		gpio_direction_output(npcs_pin, !(spi->mode & SPI_CS_HIGH));
 	} else {
-		unsigned long		flags;
-
-		spin_lock_irqsave(&as->lock, flags);
+		atmel_spi_lock(as);
 		if (as->stay == spi)
 			as->stay = NULL;
 		cs_deactivate(as, spi);
-		spin_unlock_irqrestore(&as->lock, flags);
+		atmel_spi_unlock(as);
 	}
 
 	asd->csr = csr;

@@ -813,7 +822,6 @@ static int atmel_spi_transfer(struct spi_device *spi, struct spi_message *msg)
 {
 	struct atmel_spi	*as;
 	struct spi_transfer	*xfer;
-	unsigned long		flags;
 	struct device		*controller = spi->master->dev.parent;
 	u8			bits;
 	struct atmel_spi_device	*asd;

@@ -878,11 +886,11 @@ static int atmel_spi_transfer(struct spi_device *spi, struct spi_message *msg)
 	msg->status = -EINPROGRESS;
 	msg->actual_length = 0;
 
-	spin_lock_irqsave(&as->lock, flags);
+	atmel_spi_lock(as);
 	list_add_tail(&msg->queue, &as->queue);
 	if (!as->current_transfer)
 		atmel_spi_next_message(spi->master);
-	spin_unlock_irqrestore(&as->lock, flags);
+	atmel_spi_unlock(as);
 
 	return 0;
 }

@@ -892,17 +900,16 @@ static void atmel_spi_cleanup(struct spi_device *spi)
 	struct atmel_spi	*as = spi_master_get_devdata(spi->master);
 	struct atmel_spi_device	*asd = spi->controller_state;
 	unsigned		gpio = (unsigned) spi->controller_data;
-	unsigned long		flags;
 
 	if (!asd)
 		return;
 
-	spin_lock_irqsave(&as->lock, flags);
+	atmel_spi_lock(as);
 	if (as->stay == spi) {
 		as->stay = NULL;
 		cs_deactivate(as, spi);
 	}
-	spin_unlock_irqrestore(&as->lock, flags);
+	atmel_spi_unlock(as);
 
 	spi->controller_state = NULL;
 	gpio_free(gpio);

index 37f54c3..4cb5f05 100644
@@ -15,6 +15,7 @@
 #include <linux/platform_device.h>
 #include <linux/delay.h>
 #include <linux/dma-mapping.h>
+#include <linux/dmaengine.h>
 #include <linux/err.h>
 #include <linux/interrupt.h>
 #include <linux/spi/spi.h>

@@ -24,6 +25,7 @@
 #include <mach/board.h>
 #include <asm/gpio.h>
 #include <mach/cpu.h>
+#include <linux/platform_data/dma-atmel.h>
 
 /* SPI register offsets */
 #define SPI_CR					0x0000

@@ -179,6 +181,27 @@
 #define spi_writel(port,reg,value) \
 	__raw_writel((value), (port)->regs + SPI_##reg)
 
+/* dma_type: the dma type supported by the spi.
+ *		dma_type = 0 (no used), = 1 (pdc), = 2 (dma)
+ */
+struct atmel_spi_data {
+	u8			dma_type;
+	struct at_dma_slave	dma_slave;
+};
+
+/* use PIO for small transfers, avoiding DMA setup/teardown overhead and
+ * cache operations; better heuristics consider wordsize and bitrate.
+ */
+#define DMA_MIN_BYTES	16
+
+struct atmel_spi_dma {
+	struct dma_chan			*chan_rx;
+	struct dma_chan			*chan_tx;
+	struct scatterlist		sgrx;
+	struct scatterlist		sgtx;
+	struct dma_async_tx_descriptor	*data_desc_rx;
+	struct dma_async_tx_descriptor	*data_desc_tx;
+};
 
 /*
  * The core SPI transfer engine just talks to a register bank to set up

@@ -198,14 +221,23 @@ struct atmel_spi {
 
 	u8			stopping;
 	struct list_head	queue;
+	struct tasklet_struct	tasklet;
 	struct spi_transfer	*current_transfer;
 	unsigned long		current_remaining_bytes;
 	struct spi_transfer	*next_transfer;
 	unsigned long		next_remaining_bytes;
 	int			done_status;
+	struct atmel_spi_data	data;
+
+	bool			use_dma;
+	bool			use_pdc;
 
+	/* scratch buffer */
 	void			*buffer;
 	dma_addr_t		buffer_dma;
+
+	/* dmaengine data */
+	struct atmel_spi_dma	dma;
 };
 
 /* Controller-specific per-slave state */

@@ -271,9 +303,9 @@ static void cs_activate(struct atmel_spi *as, struct spi_device *spi)
 		 * switches to the correct idle polarity before we
 		 * toggle the CS.
 		 */
-		spi_writel(as, CSR0, asd->csr);
-		spi_writel(as, MR, SPI_BF(PCS, 0x0e) | SPI_BIT(MODFDIS)
-				| SPI_BIT(MSTR));
+		spi_writel(as, CSR0 + 4 * spi->chip_select, asd->csr);
+		spi_writel(as, MR, SPI_BF(PCS, ~(0x01 << spi->chip_select))
+				| SPI_BIT(MODFDIS) | SPI_BIT(MSTR));
 		mr = spi_readl(as, MR);
 		gpio_set_value(asd->npcs_pin, active);
 	} else {

@@ -334,6 +366,23 @@ static void atmel_spi_unlock(struct atmel_spi *as)
 		spin_unlock_irqrestore(&as->lock, as->flags);
 }
 
+static inline bool atmel_spi_use_dma(struct atmel_spi *as,
+				struct spi_transfer *xfer)
+{
+	if ((as->use_dma) && (xfer->len >= DMA_MIN_BYTES))
+		return true;
+	else
+		return false;
+}
+
+static inline bool atmel_spi_use_pdc(struct atmel_spi *as)
+{
+	if (as->use_pdc)
+		return true;
+	else
+		return false;
+}
+
 static inline int atmel_spi_xfer_is_last(struct spi_message *msg,
 					struct spi_transfer *xfer)
 {

@@ -345,6 +394,209 @@ static inline int atmel_spi_xfer_can_be_chained(struct spi_transfer *xfer)
 	return xfer->delay_usecs == 0 && !xfer->cs_change;
 }
 
+static bool filter(struct dma_chan *chan, void *slave)
+{
+	struct	at_dma_slave *sl = slave;
+
+	if (sl->dma_dev == chan->device->dev) {
+		chan->private = sl;
+		return true;
+	} else {
+		return false;
+	}
+}
+
+static int __devinit atmel_spi_configure_dma(struct atmel_spi *as)
+{
+	struct at_dma_slave *sdata = (struct at_dma_slave *)&as->data.dma_slave;
+
+	if (sdata && sdata->dma_dev) {
+		dma_cap_mask_t mask;
+
+		/* setup DMA addresses */
+		sdata->rx_reg = (dma_addr_t)as->phybase + SPI_RDR;
+		sdata->tx_reg = (dma_addr_t)as->phybase + SPI_TDR;
+
+		/* Try to grab two DMA channels */
+		dma_cap_zero(mask);
+		dma_cap_set(DMA_SLAVE, mask);
+		as->dma.chan_tx = dma_request_channel(mask, filter, sdata);
+		if (as->dma.chan_tx)
+			as->dma.chan_rx =
+				dma_request_channel(mask, filter, sdata);
+	}
+	if (!as->dma.chan_rx || !as->dma.chan_tx) {
+		if (as->dma.chan_rx)
+			dma_release_channel(as->dma.chan_rx);
+		if (as->dma.chan_tx)
+			dma_release_channel(as->dma.chan_tx);
+		dev_err(&as->pdev->dev, "DMA channel not available, " \
+					"unable to use SPI\n");
+		return -EBUSY;
+	}
+
+	dev_info(&as->pdev->dev, "Using %s (tx) and " \
+				" %s (rx) for DMA transfers\n",
+				dma_chan_name(as->dma.chan_tx),
+				dma_chan_name(as->dma.chan_rx));
+
+	return 0;
+}
+
+static void atmel_spi_stop_dma(struct atmel_spi *as)
+{
+	if (as->dma.chan_rx)
+		as->dma.chan_rx->device->device_control(as->dma.chan_rx,
+							DMA_TERMINATE_ALL, 0);
+	if (as->dma.chan_tx)
+		as->dma.chan_tx->device->device_control(as->dma.chan_tx,
+							DMA_TERMINATE_ALL, 0);
+}
+
+static void atmel_spi_release_dma(struct atmel_spi *as)
+{
+	if (as->dma.chan_rx)
+		dma_release_channel(as->dma.chan_rx);
+	if (as->dma.chan_tx)
+		dma_release_channel(as->dma.chan_tx);
+}
+
+/* This function is called by the DMA driver from tasklet context */
+static void dma_callback(void *data)
+{
+	struct spi_master	*master = data;
+	struct atmel_spi	*as = spi_master_get_devdata(master);
+
+	/* trigger SPI tasklet */
+	tasklet_schedule(&as->tasklet);
+}
+
+/*
+ * Next transfer using PIO.
+ * lock is held, spi tasklet is blocked
+ */
+static void atmel_spi_next_xfer_pio(struct spi_master *master,
+				struct spi_transfer *xfer)
+{
+	struct atmel_spi	*as = spi_master_get_devdata(master);
+
+	dev_vdbg(master->dev.parent, "atmel_spi_next_xfer_pio\n");
+
+	as->current_remaining_bytes = xfer->len;
+
+	/* Make sure data is not remaining in RDR */
+	spi_readl(as, RDR);
+	while (spi_readl(as, SR) & SPI_BIT(RDRF)) {
+		spi_readl(as, RDR);
+		cpu_relax();
+	}
+
+	if (xfer->tx_buf)
+		spi_writel(as, TDR, *(u8 *)(xfer->tx_buf));
+	else
+		spi_writel(as, TDR, 0);
+
+	dev_dbg(master->dev.parent,
+		"  start pio xfer %p: len %u tx %p rx %p\n",
+		xfer, xfer->len, xfer->tx_buf, xfer->rx_buf);
+
+	/* Enable relevant interrupts */
+	spi_writel(as, IER, SPI_BIT(RDRF) | SPI_BIT(OVRES));
+}
+
+/*
+ * Submit next transfer for DMA.
+ * lock is held, spi tasklet is blocked
+ */
+static int atmel_spi_next_xfer_dma(struct spi_master *master,
+				struct spi_transfer *xfer)
+{
+	struct atmel_spi	*as = spi_master_get_devdata(master);
+	struct dma_chan		*rxchan = as->dma.chan_rx;
+	struct dma_chan		*txchan = as->dma.chan_tx;
+	struct dma_async_tx_descriptor *rxdesc;
+	struct dma_async_tx_descriptor *txdesc;
+	dma_cookie_t		cookie;
+
+	dev_vdbg(master->dev.parent, "atmel_spi_next_xfer_dma\n");
+
+	/* Check that the channels are available */
+	if (!rxchan || !txchan)
+		return -ENODEV;
+
+	/* release lock for DMA operations */
+	atmel_spi_unlock(as);
+
+	/* prepare the RX dma transfer */
+	sg_init_table(&as->dma.sgrx, 1);
+	sg_dma_len(&as->dma.sgrx) = xfer->len;
+	if (xfer->rx_buf)
+		as->dma.sgrx.dma_address = xfer->rx_dma;
+	else
+		as->dma.sgrx.dma_address = as->buffer_dma;
+
+	/* prepare the TX dma transfer */
+	sg_init_table(&as->dma.sgtx, 1);
+	sg_dma_len(&as->dma.sgtx) = xfer->len;
+	if (xfer->tx_buf) {
+		as->dma.sgtx.dma_address = xfer->tx_dma;
+	} else {
+		as->dma.sgtx.dma_address = as->buffer_dma;
+		memset(as->buffer, 0, xfer->len);
+	}
+
+	/* Send both scatterlists */
+	rxdesc = rxchan->device->device_prep_slave_sg(rxchan,
+					&as->dma.sgrx,
+					1,
+					DMA_FROM_DEVICE,
+					DMA_PREP_INTERRUPT | DMA_CTRL_ACK,
+					NULL);
+	if (!rxdesc)
+		goto err_dma;
+
+	txdesc = txchan->device->device_prep_slave_sg(txchan,
+					&as->dma.sgtx,
+					1,
+					DMA_TO_DEVICE,
+					DMA_PREP_INTERRUPT | DMA_CTRL_ACK,
+					NULL);
+	if (!txdesc)
+		goto err_dma;
+
+	dev_dbg(master->dev.parent,
+		"  start dma xfer %p: len %u tx %p/%08x rx %p/%08x\n",
+		xfer, xfer->len, xfer->tx_buf, xfer->tx_dma,
+		xfer->rx_buf, xfer->rx_dma);
+
+	/* Enable relevant interrupts */
+	spi_writel(as, IER, SPI_BIT(OVRES));
+
+	/* Put the callback on the RX transfer only, that should finish last */
+	rxdesc->callback = dma_callback;
+	rxdesc->callback_param = master;
+
+	/* Submit and fire RX and TX with TX last so we're ready to read! */
+	cookie = rxdesc->tx_submit(rxdesc);
+	if (dma_submit_error(cookie))
+		goto err_dma;
+	cookie = txdesc->tx_submit(txdesc);
+	if (dma_submit_error(cookie))
+		goto err_dma;
+	rxchan->device->device_issue_pending(rxchan);
+	txchan->device->device_issue_pending(txchan);
+
+	/* take back lock */
+	atmel_spi_lock(as);
+	return 0;
+
+err_dma:
+	spi_writel(as, IDR, SPI_BIT(OVRES));
+	atmel_spi_stop_dma(as);
+	atmel_spi_lock(as);
+	return -ENOMEM;
+}
+
 static void atmel_spi_next_xfer_data(struct spi_master *master,
 				struct spi_transfer *xfer,
 				dma_addr_t *tx_dma,

@@ -377,10 +629,10 @@ static void atmel_spi_next_xfer_data(struct spi_master *master,
 }
 
 /*
- * Submit next transfer for DMA.
+ * Submit next transfer for PDC.
  * lock is held, spi irq is blocked
  */
-static void atmel_spi_next_xfer(struct spi_master *master,
+static void atmel_spi_next_xfer_pdc(struct spi_master *master,
 				struct spi_message *msg)
 {
 	struct atmel_spi	*as = spi_master_get_devdata(master);

@@ -477,6 +729,44 @@ static void atmel_spi_next_xfer(struct spi_master *master,
 	spi_writel(as, PTCR, SPI_BIT(TXTEN) | SPI_BIT(RXTEN));
 }
 
+/*
+ * Choose way to submit next transfer and start it.
+ * lock is held, spi tasklet is blocked
+ */
+static void atmel_spi_next_xfer(struct spi_master *master,
+				struct spi_message *msg)
+{
+	struct atmel_spi	*as = spi_master_get_devdata(master);
+	struct spi_transfer	*xfer;
+
+	dev_vdbg(&msg->spi->dev, "atmel_spi_next_xfer\n");
+
+	if (as->use_pdc)
+		atmel_spi_next_xfer_pdc(master, msg);
+	else {
+		if (!as->current_transfer)
+			xfer = list_entry(msg->transfers.next,
+				struct spi_transfer, transfer_list);
+		else
+			xfer = list_entry(
+				as->current_transfer->transfer_list.next,
+				struct spi_transfer, transfer_list);
+
+		as->current_transfer = xfer;
+
+		/* quick (and *really* not optimal) workaround for DMA BUG */
+		if (atmel_spi_use_dma(as, xfer)) {
+			if (!atmel_spi_next_xfer_dma(master, xfer))
+				return;
+			else
+				dev_err(&msg->spi->dev, "unable to use DMA, fallback to PIO\n");
+		}
+
+		/* use PIO if xfer is short or error appened using DMA */
+		atmel_spi_next_xfer_pio(master, xfer);
+	}
+}
+
 static void atmel_spi_next_message(struct spi_master *master)
 {
 	struct atmel_spi	*as = spi_master_get_devdata(master);

@@ -554,6 +844,11 @@ static void atmel_spi_dma_unmap_xfer(struct spi_master *master,
 				 xfer->len, DMA_FROM_DEVICE);
 }
 
+static void atmel_spi_disable_pdc_transfer(struct atmel_spi *as)
+{
+	spi_writel(as, PTCR, SPI_BIT(RXTDIS) | SPI_BIT(TXTDIS));
+}
+
 static void
 atmel_spi_msg_done(struct spi_master *master, struct atmel_spi *as,
 		struct spi_message *msg, int stay)

@@ -579,19 +874,170 @@ atmel_spi_msg_done(struct spi_master *master, struct atmel_spi *as,
 	as->done_status = 0;
 
 	/* continue if needed */
-	if (list_empty(&as->queue) || as->stopping)
-		spi_writel(as, PTCR, SPI_BIT(RXTDIS) | SPI_BIT(TXTDIS));
-	else
+	if (list_empty(&as->queue) || as->stopping) {
+		if (as->use_pdc)
+			atmel_spi_disable_pdc_transfer(as);
+	} else
 		atmel_spi_next_message(master);
 }
 
-static irqreturn_t
-atmel_spi_interrupt(int irq, void *dev_id)
+/* Called from IRQ
+ * lock is held
+ *
+ * Must update "current_remaining_bytes" to keep track of data
+ * to transfer.
+ */
+static void
+atmel_spi_pump_pio_data(struct atmel_spi *as, struct spi_transfer *xfer)
 {
-	struct spi_master	*master = dev_id;
+	u8		*txp;
+	u8		*rxp;
+	unsigned long	xfer_pos = xfer->len - as->current_remaining_bytes;
+
+	if (xfer->rx_buf) {
+		rxp = ((u8 *)xfer->rx_buf) + xfer_pos;
+		*rxp = spi_readl(as, RDR);
+	} else {
+		spi_readl(as, RDR);
+	}
+
+	as->current_remaining_bytes--;
+
+	if (as->current_remaining_bytes) {
+		if (xfer->tx_buf) {
+			txp = ((u8 *)xfer->tx_buf) + xfer_pos + 1;
+			spi_writel(as, TDR, *txp);
+		} else {
+			spi_writel(as, TDR, 0);
+		}
+	}
+}
+
+/* Tasklet
+ * Called from DMA callback + pio transfer and overrun IRQ.
+ */
+static void atmel_spi_tasklet_func(unsigned long data)
+{
+	struct spi_master	*master = (struct spi_master *)data;
 	struct atmel_spi	*as = spi_master_get_devdata(master);
 	struct spi_message	*msg;
 	struct spi_transfer	*xfer;
+
+	dev_vdbg(master->dev.parent, "atmel_spi_tasklet_func\n");
+
+	atmel_spi_lock(as);
+
+	xfer = as->current_transfer;
+
+	if (xfer == NULL)
+		/* already been there */
+		goto tasklet_out;
+
+	msg = list_entry(as->queue.next, struct spi_message, queue);
+
+	if (as->done_status < 0) {
+		/* error happened (overrun) */
+		if (atmel_spi_use_dma(as, xfer))
+			atmel_spi_stop_dma(as);
+	} else {
+		/* only update length if no error */
+		msg->actual_length += xfer->len;
+	}
+
+	if (atmel_spi_use_dma(as, xfer)) {
+		if (!msg->is_dma_mapped)
+			atmel_spi_dma_unmap_xfer(master, xfer);
+	}
+
+	if (xfer->delay_usecs)
+		udelay(xfer->delay_usecs);
+
+	if (atmel_spi_xfer_is_last(msg, xfer) || as->done_status < 0) {
+		/* report completed (or erroneous) message */
+		atmel_spi_msg_done(master, as, msg, xfer->cs_change);
+	} else {
+		if (xfer->cs_change) {
+			cs_deactivate(as, msg->spi);
+			udelay(1);
+			cs_activate(as, msg->spi);
+		}
+
+		/*
+		 * Not done yet. Submit the next transfer.
+		 *
+		 * FIXME handle protocol options for xfer
+		 */
+		atmel_spi_next_xfer(master, msg);
+	}
+
+tasklet_out:
+	atmel_spi_unlock(as);
+}
+
+static int atmel_spi_interrupt_dma(struct atmel_spi *as,
+				struct spi_master *master)
+{
+	u32			status, pending, imr;
+	struct spi_transfer	*xfer;
+	int			ret = IRQ_NONE;
+
+	imr = spi_readl(as, IMR);
+	status = spi_readl(as, SR);
+	pending = status & imr;
+
+	if (pending & SPI_BIT(OVRES)) {
+		ret = IRQ_HANDLED;
+		spi_writel(as, IDR, SPI_BIT(OVRES));
+		dev_warn(master->dev.parent, "overrun\n");
+
+		/*
+		 * When we get an overrun, we disregard the current
+		 * transfer. Data will not be copied back from any
+		 * bounce buffer and msg->actual_len will not be
+		 * updated with the last xfer.
+		 *
+		 * We will also not process any remaning transfers in
+		 * the message.
+		 *
+		 * All actions are done in tasklet with done_status indication
+		 */
+		as->done_status = -EIO;
+		smp_wmb();
+
+		/* Clear any overrun happening while cleaning up */
+		spi_readl(as, SR);
+
+		tasklet_schedule(&as->tasklet);
+
+	} else if (pending & SPI_BIT(RDRF)) {
+		atmel_spi_lock(as);
+
+		if (as->current_remaining_bytes) {
+			ret = IRQ_HANDLED;
+			xfer = as->current_transfer;
+			atmel_spi_pump_pio_data(as, xfer);
+			if (!as->current_remaining_bytes) {
+				/* no more data to xfer, kick tasklet */
+				spi_writel(as, IDR, pending);
+				tasklet_schedule(&as->tasklet);
+			}
+		}
+
+		atmel_spi_unlock(as);
+	} else {
+		WARN_ONCE(pending, "IRQ not handled, pending = %x\n", pending);
+		ret = IRQ_HANDLED;
+		spi_writel(as, IDR, pending);
+	}
+
+	return ret;
+}
+
+static int atmel_spi_interrupt_pdc(struct atmel_spi *as,
+				struct spi_master *master)
+{
+	struct spi_message	*msg;
+	struct spi_transfer	*xfer;
 	u32			status, pending, imr;
 	int			ret = IRQ_NONE;
 

@@ -703,6 +1149,27 @@ atmel_spi_interrupt(int irq, void *dev_id)
 	return ret;
 }
 
+/* Interrupt
+ *
+ * No need for locking in this Interrupt handler: done_status is the
+ * only information modified. What we need is the update of this field
+ * before tasklet runs. This is ensured by using barrier.
+ */
+static irqreturn_t
+atmel_spi_interrupt(int irq, void *dev_id)
+{
+	struct spi_master	*master = dev_id;
+	struct atmel_spi	*as = spi_master_get_devdata(master);
+	int ret;
+
+	if (as->use_pdc)
+		ret = atmel_spi_interrupt_pdc(as, master);
+	else
+		ret = atmel_spi_interrupt_dma(as, master);
+
+	return ret;
+}
+
 static int atmel_spi_setup(struct spi_device *spi)
 {
 	struct atmel_spi	*as;

@@ -861,13 +1328,11 @@ static int atmel_spi_transfer(struct spi_device *spi, struct spi_message *msg)
 
 		/*
 		 * DMA map early, for performance (empties dcache ASAP) and
-		 * better fault reporting.  This is a DMA-only driver.
-		 *
-		 * NOTE that if dma_unmap_single() ever starts to do work on
-		 * platforms supported by this driver, we would need to clean
-		 * up mappings for previously-mapped transfers.
+		 * better fault reporting.
 		 */
-		if (!msg->is_dma_mapped) {
+		if (!msg->is_dma_mapped
+			&& (atmel_spi_use_dma(as, xfer)
+				|| atmel_spi_use_pdc(as))) {
 			if (atmel_spi_dma_map_xfer(as, xfer) < 0)
 				return -ENOMEM;
 		}

@@ -968,6 +1433,8 @@ static int __devinit atmel_spi_probe(struct platform_device *pdev)
 
 	spin_lock_init(&as->lock);
 	INIT_LIST_HEAD(&as->queue);
+	tasklet_init(&as->tasklet, atmel_spi_tasklet_func,
+					(unsigned long)master);
 	as->pdev = pdev;
 	as->regs = ioremap(regs->start, resource_size(regs));
 	if (!as->regs)

@@ -986,7 +1453,16 @@ static int __devinit atmel_spi_probe(struct platform_device *pdev)
 	spi_writel(as, CR, SPI_BIT(SWRST));
 	spi_writel(as, CR, SPI_BIT(SWRST)); /* AT91SAM9263 Rev B workaround */
 	spi_writel(as, MR, SPI_BIT(MSTR) | SPI_BIT(MODFDIS));
-	spi_writel(as, PTCR, SPI_BIT(RXTDIS) | SPI_BIT(TXTDIS));
+
+	as->use_dma = false;
+	as->use_pdc = false;
+
+	if (as->data.dma_type == 2) {
+		if (atmel_spi_configure_dma(as) == 0)
+			as->use_dma = true;
+	} else if (as->data.dma_type == 1)
+		as->use_pdc = true;
+
 	spi_writel(as, CR, SPI_BIT(SPIEN));
 
 	/* go! */

@@ -995,11 +1471,14 @@ static int __devinit atmel_spi_probe(struct platform_device *pdev)
 
 	ret = spi_register_master(master);
 	if (ret)
-		goto out_reset_hw;
+		goto out_free_dma;
 
 	return 0;
 
-out_reset_hw:
+out_free_dma:
+	if (as->use_dma)
+		atmel_spi_release_dma(as);
+
 	spi_writel(as, CR, SPI_BIT(SWRST));
 	spi_writel(as, CR, SPI_BIT(SWRST)); /* AT91SAM9263 Rev B workaround */
 	clk_disable(clk);

@@ -1007,6 +1486,7 @@ out_reset_hw:
 out_unmap_regs:
 	iounmap(as->regs);
 out_free_buffer:
+	tasklet_kill(&as->tasklet);
 	dma_free_coherent(&pdev->dev, BUFFER_SIZE, as->buffer,
 			as->buffer_dma);
 out_free:

@@ -1025,6 +1505,11 @@ static int __devexit atmel_spi_remove(struct platform_device *pdev)
 	/* reset the hardware and block queue progress */
 	spin_lock_irq(&as->lock);
 	as->stopping = 1;
+	if (as->use_dma) {
+		atmel_spi_stop_dma(as);
+		atmel_spi_release_dma(as);
+	}
+
 	spi_writel(as, CR, SPI_BIT(SWRST));
 	spi_writel(as, CR, SPI_BIT(SWRST)); /* AT91SAM9263 Rev B workaround */
 	spi_readl(as, SR);

@@ -1033,13 +1518,16 @@ static int __devexit atmel_spi_remove(struct platform_device *pdev)
 	/* Terminate remaining queued transfers */
 	list_for_each_entry(msg, &as->queue, queue) {
 		list_for_each_entry(xfer, &msg->transfers, transfer_list) {
-			if (!msg->is_dma_mapped)
+			if (!msg->is_dma_mapped
+				&& (atmel_spi_use_dma(as, xfer)
+					|| atmel_spi_use_pdc(as)))
 				atmel_spi_dma_unmap_xfer(master, xfer);
 		}
 		msg->status = -ESHUTDOWN;
 		msg->complete(msg->context);
 	}
 
+	tasklet_kill(&as->tasklet);
 	dma_free_coherent(&pdev->dev, BUFFER_SIZE, as->buffer,
 			as->buffer_dma);
 

index 4cb5f05..00e729b 100644
@@ -409,14 +409,21 @@ static bool filter(struct dma_chan *chan, void *slave)
 static int __devinit atmel_spi_configure_dma(struct atmel_spi *as)
 {
 	struct at_dma_slave *sdata = (struct at_dma_slave *)&as->data.dma_slave;
+	struct dma_slave_config	slave_config;
+	int err;
+
+	memset(&slave_config, 0, sizeof(slave_config));
+	slave_config.dst_addr = (dma_addr_t)as->phybase + SPI_TDR;
+	slave_config.src_addr = (dma_addr_t)as->phybase + SPI_RDR;
+	slave_config.dst_addr_width = DMA_SLAVE_BUSWIDTH_1_BYTE;
+	slave_config.src_addr_width = DMA_SLAVE_BUSWIDTH_1_BYTE;
+	slave_config.src_maxburst = 1;
+	slave_config.dst_maxburst = 1;
+	slave_config.device_fc = false;
 
 	if (sdata && sdata->dma_dev) {
 		dma_cap_mask_t mask;
 
-		/* setup DMA addresses */
-		sdata->rx_reg = (dma_addr_t)as->phybase + SPI_RDR;
-		sdata->tx_reg = (dma_addr_t)as->phybase + SPI_TDR;
-
 		/* Try to grab two DMA channels */
 		dma_cap_zero(mask);
 		dma_cap_set(DMA_SLAVE, mask);

@@ -426,13 +433,26 @@ static int __devinit atmel_spi_configure_dma(struct atmel_spi *as)
 				dma_request_channel(mask, filter, sdata);
 	}
 	if (!as->dma.chan_rx || !as->dma.chan_tx) {
-		if (as->dma.chan_rx)
-			dma_release_channel(as->dma.chan_rx);
-		if (as->dma.chan_tx)
-			dma_release_channel(as->dma.chan_tx);
 		dev_err(&as->pdev->dev, "DMA channel not available, " \
 					"unable to use SPI\n");
-		return -EBUSY;
+		err = -EBUSY;
+		goto error;
+	}
+
+	slave_config.direction = DMA_TO_DEVICE;
+	if (dmaengine_slave_config(as->dma.chan_tx, &slave_config)) {
+		dev_err(&as->pdev->dev,
+			"failed to configure tx dma channel\n");
+		err = -EINVAL;
+		goto error;
+	}
+
+	slave_config.direction = DMA_FROM_DEVICE;
+	if (dmaengine_slave_config(as->dma.chan_rx, &slave_config)) {
+		dev_err(&as->pdev->dev,
+			"failed to configure rx dma channel\n");
+		err = -EINVAL;
+		goto error;
 	}
 
 	dev_info(&as->pdev->dev, "Using %s (tx) and " \

@@ -441,6 +461,12 @@ static int __devinit atmel_spi_configure_dma(struct atmel_spi *as)
 				dma_chan_name(as->dma.chan_rx));
 
 	return 0;
+error:
+	if (as->dma.chan_rx)
+		dma_release_channel(as->dma.chan_rx);
+	if (as->dma.chan_tx)
+		dma_release_channel(as->dma.chan_tx);
+	return err;
 }
 
 static void atmel_spi_stop_dma(struct atmel_spi *as)

index 00e729b..0007a53 100644
@@ -518,13 +518,17 @@ static void atmel_spi_next_xfer_pio(struct spi_master *master,
 	}
 
 	if (xfer->tx_buf)
-		spi_writel(as, TDR, *(u8 *)(xfer->tx_buf));
+		if (xfer->bits_per_word > 8)
+			spi_writel(as, TDR, *(u16 *)(xfer->tx_buf));
+		else
+			spi_writel(as, TDR, *(u8 *)(xfer->tx_buf));
 	else
 		spi_writel(as, TDR, 0);
 
 	dev_dbg(master->dev.parent,
-		"  start pio xfer %p: len %u tx %p rx %p\n",
-		xfer, xfer->len, xfer->tx_buf, xfer->rx_buf);
+		"  start pio xfer %p: len %u tx %p rx %p bitpw %d\n",
+		xfer, xfer->len, xfer->tx_buf, xfer->rx_buf,
+		xfer->bits_per_word);
 
 	/* Enable relevant interrupts */
 	spi_writel(as, IER, SPI_BIT(RDRF) | SPI_BIT(OVRES));

@@ -918,21 +922,39 @@ atmel_spi_pump_pio_data(struct atmel_spi *as, struct spi_transfer *xfer)
 {
 	u8		*txp;
 	u8		*rxp;
+	u16		*txp16;
+	u16		*rxp16;
 	unsigned long	xfer_pos = xfer->len - as->current_remaining_bytes;
 
 	if (xfer->rx_buf) {
-		rxp = ((u8 *)xfer->rx_buf) + xfer_pos;
-		*rxp = spi_readl(as, RDR);
+		if (xfer->bits_per_word > 8) {
+			rxp16 = (u16 *)(((u8 *)xfer->rx_buf) + xfer_pos);
+			*rxp16 = spi_readl(as, RDR);
+		} else {
+			rxp = ((u8 *)xfer->rx_buf) + xfer_pos;
+			*rxp = spi_readl(as, RDR);
+		}
 	} else {
 		spi_readl(as, RDR);
 	}
-
-	as->current_remaining_bytes--;
+	if (xfer->bits_per_word > 8) {
+		as->current_remaining_bytes -= 2;
+		if (as->current_remaining_bytes < 0)
+			as->current_remaining_bytes = 0;
+	} else {
+		as->current_remaining_bytes--;
+	}
 
 	if (as->current_remaining_bytes) {
 		if (xfer->tx_buf) {
-			txp = ((u8 *)xfer->tx_buf) + xfer_pos + 1;
-			spi_writel(as, TDR, *txp);
+			if (xfer->bits_per_word > 8) {
+				txp16 = (u16 *)(((u8 *)xfer->tx_buf)
+							+ xfer_pos + 2);
+				spi_writel(as, TDR, *txp16);
+			} else {
+				txp = ((u8 *)xfer->tx_buf) + xfer_pos + 1;
+				spi_writel(as, TDR, *txp);
+			}
 		} else {
 			spi_writel(as, TDR, 0);
 		}

@@ -1345,6 +1367,12 @@ static int atmel_spi_transfer(struct spi_device *spi, struct spi_message *msg)
 				return -ENOPROTOOPT;
 			}
 		}
+		if (xfer->bits_per_word > 8) {
+			if (xfer->len % 2) {
+				dev_dbg(&spi->dev, "buffer len should be 16 bits aligned\n");
+				return -EINVAL;
+			}
+		}
 
 		/* FIXME implement these protocol options!! */
 		if (xfer->speed_hz) {

index 0007a53..7a3613d 100644
@@ -249,6 +249,8 @@ struct atmel_spi_device {
 #define BUFFER_SIZE		PAGE_SIZE
 #define INVALID_DMA_ADDRESS	0xffffffff
 
+static struct dma_slave_config slave_config;
+
 /*
  * Version 2 of the SPI controller has
  *  - CR.LASTXFER

@@ -406,17 +408,41 @@ static bool filter(struct dma_chan *chan, void *slave)
 	}
 }
 
+static int atmel_spi_set_dma_xfer_width(struct atmel_spi *as, u8 bits_per_word)
+{
+	int err = 0;
+
+	if (bits_per_word > 8) {
+		slave_config.dst_addr_width = DMA_SLAVE_BUSWIDTH_2_BYTES;
+		slave_config.src_addr_width = DMA_SLAVE_BUSWIDTH_2_BYTES;
+	} else {
+		slave_config.dst_addr_width = DMA_SLAVE_BUSWIDTH_1_BYTE;
+		slave_config.src_addr_width = DMA_SLAVE_BUSWIDTH_1_BYTE;
+	}
+
+	slave_config.direction = DMA_TO_DEVICE;
+	if (dmaengine_slave_config(as->dma.chan_tx, &slave_config)) {
+		dev_err(&as->pdev->dev,
+			"failed to configure tx dma channel\n");
+		err = -EINVAL;
+	}
+
+	slave_config.direction = DMA_FROM_DEVICE;
+	if (dmaengine_slave_config(as->dma.chan_rx, &slave_config)) {
+		dev_err(&as->pdev->dev,
+			"failed to configure rx dma channel\n");
+		err = -EINVAL;
+	}
+	return err;
+}
+
 static int __devinit atmel_spi_configure_dma(struct atmel_spi *as)
 {
 	struct at_dma_slave *sdata = (struct at_dma_slave *)&as->data.dma_slave;
-	struct dma_slave_config	slave_config;
 	int err;
 
-	memset(&slave_config, 0, sizeof(slave_config));
 	slave_config.dst_addr = (dma_addr_t)as->phybase + SPI_TDR;
 	slave_config.src_addr = (dma_addr_t)as->phybase + SPI_RDR;
-	slave_config.dst_addr_width = DMA_SLAVE_BUSWIDTH_1_BYTE;
-	slave_config.src_addr_width = DMA_SLAVE_BUSWIDTH_1_BYTE;
 	slave_config.src_maxburst = 1;
 	slave_config.dst_maxburst = 1;
 	slave_config.device_fc = false;

@@ -439,21 +465,9 @@ static int __devinit atmel_spi_configure_dma(struct atmel_spi *as)
 		goto error;
 	}
 
-	slave_config.direction = DMA_TO_DEVICE;
-	if (dmaengine_slave_config(as->dma.chan_tx, &slave_config)) {
-		dev_err(&as->pdev->dev,
-			"failed to configure tx dma channel\n");
-		err = -EINVAL;
+	err = atmel_spi_set_dma_xfer_width(as, 8);
+	if (err)
 		goto error;
-	}
-
-	slave_config.direction = DMA_FROM_DEVICE;
-	if (dmaengine_slave_config(as->dma.chan_rx, &slave_config)) {
-		dev_err(&as->pdev->dev,
-			"failed to configure rx dma channel\n");
-		err = -EINVAL;
-		goto error;
-	}
 
 	dev_info(&as->pdev->dev, "Using %s (tx) and " \
 				" %s (rx) for DMA transfers\n",

@@ -575,6 +589,9 @@ static int atmel_spi_next_xfer_dma(struct spi_master *master,
 		memset(as->buffer, 0, xfer->len);
 	}
 
+	if (atmel_spi_set_dma_xfer_width(as, xfer->bits_per_word))
+		goto err_dma;
+
 	/* Send both scatterlists */
 	rxdesc = rxchan->device->device_prep_slave_sg(rxchan,
 					&as->dma.sgrx,

index 7a3613d..0cc347c 100644
@@ -20,6 +20,7 @@
 #include <linux/interrupt.h>
 #include <linux/spi/spi.h>
 #include <linux/slab.h>
+#include <linux/of.h>
 
 #include <asm/io.h>
 #include <mach/board.h>

@@ -1242,7 +1243,7 @@ static int atmel_spi_setup(struct spi_device *spi)
 	u32			scbr, csr;
 	unsigned int		bits = spi->bits_per_word;
 	unsigned long		bus_hz;
-	unsigned int		npcs_pin;
+	int			npcs_pin;
 	int			ret;
 
 	as = spi_master_get_devdata(spi->master);

@@ -1314,7 +1315,9 @@ static int atmel_spi_setup(struct spi_device *spi)
 	csr |= SPI_BF(DLYBCT, 0);
 
 	/* chipselect must have been muxed as GPIO (e.g. in board setup) */
-	npcs_pin = (unsigned int)spi->controller_data;
+	if (!gpio_is_valid(spi->cs_gpio))
+		spi->cs_gpio = (int)spi->controller_data;
+	npcs_pin = spi->cs_gpio;
 	asd = spi->controller_state;
 	if (!asd) {
 		asd = kzalloc(sizeof(struct atmel_spi_device), GFP_KERNEL);

@@ -1435,7 +1438,7 @@ static void atmel_spi_cleanup(struct spi_device *spi)
 {
 	struct atmel_spi	*as = spi_master_get_devdata(spi->master);
 	struct atmel_spi_device	*asd = spi->controller_state;
-	unsigned		gpio = (unsigned) spi->controller_data;
+	unsigned		gpio = spi->cs_gpio;
 
 	if (!asd)
 		return;

@@ -1485,7 +1488,8 @@ static int __devinit atmel_spi_probe(struct platform_device *pdev)
 	master->mode_bits = SPI_CPOL | SPI_CPHA | SPI_CS_HIGH;
 
 	master->bus_num = pdev->id;
-	master->num_chipselect = 4;
+	master->dev.of_node = pdev->dev.of_node;
+	master->num_chipselect = master->dev.of_node ? 0 : 4;
 	master->setup = atmel_spi_setup;
 	master->transfer = atmel_spi_transfer;
 	master->cleanup = atmel_spi_cleanup;

@@ -1637,11 +1641,20 @@ static int atmel_spi_resume(struct platform_device *pdev)
 #define	atmel_spi_resume	NULL
 #endif
 
+#if defined(CONFIG_OF)
+static const struct of_device_id atmel_spi_dt_ids[] = {
+	{ .compatible = "atmel,at91rm9200-spi" },
+	{ /* sentinel */ }
+};
+
+MODULE_DEVICE_TABLE(of, atmel_spi_dt_ids);
+#endif
 
 static struct platform_driver atmel_spi_driver = {
 	.driver		= {
 		.name	= "atmel_spi",
 		.owner	= THIS_MODULE,
+		.of_match_table	= of_match_ptr(atmel_spi_dt_ids),
 	},
 	.suspend	= atmel_spi_suspend,
 	.resume		= atmel_spi_resume,

index 0cc347c..568df5b 100644
@@ -187,6 +187,7 @@
  */
 struct atmel_spi_data {
 	u8			dma_type;
+	u8			version;
 	struct at_dma_slave	dma_slave;
 };
 

@@ -264,9 +265,20 @@ static struct dma_slave_config slave_config;
  * register, but I haven't checked that it exists on all chips, and
  * this is cheaper anyway.
  */
-static bool atmel_spi_is_v2(void)
+static bool atmel_spi_is_v1(struct atmel_spi *as)
 {
-	return !cpu_is_at91rm9200();
+	if (as->data.version == 1)
+		return true;
+	else
+		return false;
+}
+
+static bool atmel_spi_is_v2(struct atmel_spi *as)
+{
+	if (as->data.version == 2)
+		return true;
+	else
+		return false;
 }
 
 /*

@@ -300,7 +312,7 @@ static void cs_activate(struct atmel_spi *as, struct spi_device *spi)
 	unsigned active = spi->mode & SPI_CS_HIGH;
 	u32 mr;
 
-	if (atmel_spi_is_v2()) {
+	if (atmel_spi_is_v2(as)) {
 		/*
 		 * Always use CSR0. This ensures that the clock
 		 * switches to the correct idle polarity before we

@@ -355,7 +367,7 @@ static void cs_deactivate(struct atmel_spi *as, struct spi_device *spi)
 			asd->npcs_pin, active ? " (low)" : "",
 			mr);
 
-	if (atmel_spi_is_v2() || spi->chip_select != 0)
+	if (atmel_spi_is_v2(as) || spi->chip_select != 0)
 		gpio_set_value(asd->npcs_pin, !active);
 }
 

@@ -1266,7 +1278,7 @@ static int atmel_spi_setup(struct spi_device *spi)
 	}
 
 	/* see notes above re chipselect */
-	if (!atmel_spi_is_v2()
+	if (atmel_spi_is_v1(as)
 			&& spi->chip_select == 0
 			&& (spi->mode & SPI_CS_HIGH)) {
 		dev_dbg(&spi->dev, "setup: can't be active-high\n");

@@ -1275,7 +1287,7 @@ static int atmel_spi_setup(struct spi_device *spi)
 
 	/* v1 chips start out at half the peripheral bus speed. */
 	bus_hz = clk_get_rate(as->clk);
-	if (!atmel_spi_is_v2())
+	if (atmel_spi_is_v1(as))
 		bus_hz /= 2;
 
 	if (spi->max_speed_hz) {

@@ -1347,7 +1359,7 @@ static int atmel_spi_setup(struct spi_device *spi)
 		"setup: %lu Hz bpw %u mode 0x%x -> csr%d %08x\n",
 		bus_hz / scbr, bits, spi->mode, spi->chip_select, csr);
 
-	if (!atmel_spi_is_v2())
+	if (atmel_spi_is_v1(as))
 		spi_writel(as, CSR0 + 4 * spi->chip_select, csr);
 
 	return 0;

index 568df5b..791800e 100644
@@ -1467,6 +1467,30 @@ static void atmel_spi_cleanup(struct spi_device *spi)
 	kfree(asd);
 }
 
+static int of_get_atmel_spi_data(struct device_node *np, struct atmel_spi *as)
+{
+	const __be32	*val;
+
+	val = of_get_property(np, "dma_type", NULL);
+	if (!val) {
+		pr_err("%s: have no 'dma_type' property\n",
+						np->full_name);
+		return -EINVAL;
+	}
+
+	as->data.dma_type = be32_to_cpup(val);
+
+	val = of_get_property(np, "version", NULL);
+	if (!val) {
+		pr_err("%s: have no 'version' property\n", np->full_name);
+		return -EINVAL;
+	}
+
+	as->data.version = be32_to_cpup(val);
+
+	return 0;
+}
+
 /*-------------------------------------------------------------------------*/
 
 static int __devinit atmel_spi_probe(struct platform_device *pdev)

@@ -1535,6 +1559,10 @@ static int __devinit atmel_spi_probe(struct platform_device *pdev)
 	if (ret)
 		goto out_unmap_regs;
 
+	ret = of_get_atmel_spi_data(pdev->dev.of_node, as);
+	if (ret)
+		goto out_unmap_regs;
+
 	/* Initialize the hardware */
 	clk_enable(clk);
 	spi_writel(as, CR, SPI_BIT(SWRST));

index f820261..6c217eb 100644
@@ -230,6 +230,8 @@ static struct clk_lookup periph_clocks_lookups[] = {
 	CLKDEV_CON_DEV_ID("t1_clk", "fffdc000.timer", &tc4_clk),
 	CLKDEV_CON_DEV_ID("t2_clk", "fffdc000.timer", &tc5_clk),
 	CLKDEV_CON_DEV_ID("hclk", "500000.ohci", &ohci_clk),
+	CLKDEV_CON_DEV_ID("spi_clk", "fffc8000.spi", &spi0_clk),
+	CLKDEV_CON_DEV_ID("spi_clk", "fffcc000.spi", &spi1_clk),
 	/* fake hclk clock */
 	CLKDEV_CON_DEV_ID("hclk", "at91_ohci", &ohci_clk),
 	CLKDEV_CON_ID("pioA", &pioA_clk),

index 84af1b5..d419002 100644
@@ -256,6 +256,8 @@ static struct clk_lookup periph_clocks_lookups[] = {
 	CLKDEV_CON_DEV_ID("t0_clk", "fffd4000.timer", &tcb0_clk),
 	CLKDEV_CON_DEV_ID("hclk", "700000.ohci", &uhphs_clk),
 	CLKDEV_CON_DEV_ID("ehci_clk", "800000.ehci", &uhphs_clk),
+	CLKDEV_CON_DEV_ID("spi_clk", "fffa4000.spi", &spi0_clk),
+	CLKDEV_CON_DEV_ID("spi_clk", "fffa8000.spi", &spi1_clk),
 	CLKDEV_CON_DEV_ID(NULL, "fff84000.i2c", &twi0_clk),
 	CLKDEV_CON_DEV_ID(NULL, "fff88000.i2c", &twi1_clk),
 	/* fake hclk clock */

index 732d3d3..ea87f0e 100644
@@ -169,6 +169,8 @@ static struct clk_lookup periph_clocks_lookups[] = {
 	CLKDEV_CON_DEV_ID("t0_clk", "f8008000.timer", &tcb_clk),
 	CLKDEV_CON_DEV_ID("t0_clk", "f800c000.timer", &tcb_clk),
 	CLKDEV_CON_DEV_ID("dma_clk", "ffffec00.dma-controller", &dma_clk),
+	CLKDEV_CON_DEV_ID("spi_clk", "f0000000.spi", &spi0_clk),
+	CLKDEV_CON_DEV_ID("spi_clk", "f0004000.spi", &spi1_clk),
 	CLKDEV_CON_DEV_ID(NULL, "f8010000.i2c", &twi0_clk),
 	CLKDEV_CON_DEV_ID(NULL, "f8014000.i2c", &twi1_clk),
 	CLKDEV_CON_ID("pioA", &pioAB_clk),

index e503538..19408af 100644
@@ -231,6 +231,8 @@ static struct clk_lookup periph_clocks_lookups[] = {
 	CLKDEV_CON_DEV_ID("t0_clk", "f800c000.timer", &tcb0_clk),
 	CLKDEV_CON_DEV_ID("dma_clk", "ffffec00.dma-controller", &dma0_clk),
 	CLKDEV_CON_DEV_ID("dma_clk", "ffffee00.dma-controller", &dma1_clk),
+	CLKDEV_CON_DEV_ID("spi_clk", "f0000000.spi", &spi0_clk),
+	CLKDEV_CON_DEV_ID("spi_clk", "f0004000.spi", &spi1_clk),
 	CLKDEV_CON_DEV_ID(NULL, "f8010000.i2c", &twi0_clk),
 	CLKDEV_CON_DEV_ID(NULL, "f8014000.i2c", &twi1_clk),
 	CLKDEV_CON_DEV_ID(NULL, "f8018000.i2c", &twi2_clk),

index 03838ba..73e5fea 100644
@@ -340,6 +340,7 @@ static int m25p80_read(struct mtd_info *mtd, loff_t from, size_t len,
 	size_t *retlen, u_char *buf)
 {
 	struct m25p *flash = mtd_to_m25p(mtd);
+	u32 page_offset, page_size;
 	struct spi_transfer t[2];
 	struct spi_message m;
 

@@ -358,7 +359,6 @@ static int m25p80_read(struct mtd_info *mtd, loff_t from, size_t len,
 	spi_message_add_tail(&t[0], &m);
 
 	t[1].rx_buf = buf;
-	t[1].len = len;
 	spi_message_add_tail(&t[1], &m);
 
 	mutex_lock(&flash->lock);

@@ -379,9 +379,47 @@ static int m25p80_read(struct mtd_info *mtd, loff_t from, size_t len,
 	flash->command[0] = OPCODE_READ;
 	m25p_addr2cmd(flash, from, flash->command);
 
-	spi_sync(flash->spi, &m);
+	page_offset = from & (flash->page_size - 1);
+
+	/* do all the bytes fit onto one page? */
+	if (page_offset + len <= flash->page_size) {
+		t[1].len = len;
+
+		spi_sync(flash->spi, &m);
+
+		*retlen = m.actual_length - m25p_cmdsz(flash)
+					- FAST_READ_DUMMY_BYTE;
+
+	} else {
+		u32 i;
+
+		/* the size of data remaining on the first page */
+		page_size = flash->page_size - page_offset;
+
+		t[1].len = page_size;
+		spi_sync(flash->spi, &m);
+
+		*retlen = m.actual_length - m25p_cmdsz(flash)
+					- FAST_READ_DUMMY_BYTE;
+
+		/* write everything in flash->page_size chunks */
+		for (i = page_size; i < len; i += page_size) {
+			page_size = len - i;
+			if (page_size > flash->page_size)
+				page_size = flash->page_size;
+
+			/* write the next page to flash */
+			m25p_addr2cmd(flash, from + i, flash->command);
+
+			t[1].rx_buf = buf + i;
+			t[1].len = page_size;
+
+			spi_sync(flash->spi, &m);
 
-	*retlen = m.actual_length - m25p_cmdsz(flash) - FAST_READ_DUMMY_BYTE;
+			*retlen += m.actual_length - m25p_cmdsz(flash)
+						- FAST_READ_DUMMY_BYTE;
+		}
+	}
 
 	mutex_unlock(&flash->lock);