work in progress: SPI controller w/DMA SAM9X5

PATCH New Richard Genoud 3 participants 30 messages Aug 14, 2012 – Aug 16, 2012

Subsystems (7)

Device tree bindings Arm (32-bit) Arm sub-architectures SPI +3 more subsystems

lore

Share & tools

lore

Richard Genoud

author

2012-08-14 13:49 UTC #0 lore raw

cover work in progress: SPI controller w/DMA SAM9X5

Nicolas,
I said that I had some work done on sam9g35-ek, here is the spi part, based on your and Jean-Christophe work.
You have to enable CONFIG_ARCH_AT91SAM9X5 to get it work.

The goal of this patchset is to share the work I've done, it's not ready for upstream, but it may save some time.

It's based on 3.6-rc1 + Jean-Christophe's pinctrl patchset.

Best Regards,
Richard.

Jean-Christophe PLAGNIOL-VILLARD (3):
of: add dma-mask binding
of_spi: add generic binding support to specify cs gpio spi/atmel: add DT support

Nicolas Ferre (6):
spi/atmel_spi: trivial: change some comments spi/atmel_spi: add physical base address spi/atmel_spi: call unmapping on transfers buffers spi/atmel_spi: status information passed through controller data spi/atmel_spi: add flag to controller data for lock operations spi/atmel_spi: add dmaengine support

Richard Genoud (14):
spi-atmel: update with dmaengine interface
spi-atmel: fix __init/__devinit sections mismatch
spi-atmel: Fix spi-atmel driver to adapt to slave_config changes AT91 DMA OF support
add at91sam9x5 Kconfig ARCH/SOC link
spi-atmel: add DMA OF support
[BUG] SPI: array out of bound => no CS
[BUG] atmel-spi && DMA: OOPS if buffer > 4400 bytes
sam9x5: declare SPI clocks
spi-atmel: add sam9x5 SPI in device tree
spi-atmel: add dma support in sam9x5 device tree spi-atmel OF: complete documentation
spi-atmel: complete DMA slave OF documentation sam9x5ek DTS: enable SPI dataflash

 Documentation/devicetree/bindings/spi/spi-bus.txt  |    6 +
 .../devicetree/bindings/spi/spi_atmel.txt          |   46 ++
 arch/arm/boot/dts/at91sam9x5.dtsi                  |   42 ++
 arch/arm/boot/dts/at91sam9x5ek.dtsi                |   10 +
 arch/arm/mach-at91/Kconfig                         |    4 +
 arch/arm/mach-at91/at91sam9x5.c                    |    3 +
 arch/arm/mach-at91/include/mach/at_hdmac.h         |    1 +
 drivers/of/Kconfig                                 |    4 +
 drivers/of/Makefile                                |    1 +
 drivers/of/of_atmel.c                              |  111 ++++
 drivers/of/platform.c                              |   23 +-
 drivers/spi/Kconfig                                |    9 +
 drivers/spi/spi-atmel.c                            |  611 ++++++++++++++++++--
 drivers/spi/spi.c                                  |   55 ++-
 include/linux/of_atmel.h                           |   29 +
 include/linux/spi/spi.h                            |    3 +
 16 files changed, 914 insertions(+), 44 deletions(-)
 create mode 100644 Documentation/devicetree/bindings/spi/spi_atmel.txt
 create mode 100644 drivers/of/of_atmel.c
 create mode 100644 include/linux/of_atmel.h

Signature · 1 line

1.7.2.5

Richard Genoud

author

2012-08-14 13:49 UTC #1 lore raw

PATCH 1/23 of: add dma-mask binding

From: Jean-Christophe PLAGNIOL-VILLARD <[email protected]>

This will allow each device to specify its dma-mask for this we use the coherent_dma_mask as pointer. By default the dma-mask will be set to DMA_BIT_MASK(32).
The microblaze architecture hook is drop

Signed-off-by Jean-Christophe PLAGNIOL-VILLARD <[email protected]>

---
 drivers/of/platform.c |   23 ++++++++++++++++-------
 1 files changed, 16 insertions(+), 7 deletions(-)

drivers/of/platform.c +16 -7

index e44f8c2..11c765c 100644
@@ -121,6 +121,21 @@ 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

@@ -161,10 +176,8 @@ 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)

@@ -201,10 +214,6 @@ 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;

Signature · 1 line

1.7.2.5

Richard Genoud

author

2012-08-14 13:49 UTC #2 lore raw

PATCH 2/23 of_spi: add generic binding support to specify cs gpio

In reply to Richard Genoud

From: Jean-Christophe PLAGNIOL-VILLARD <[email protected]>

This will allow to use gpio for chip select with no modification in the driver binding

When use the cs-gpios, the gpio number will be passed via the cs_gpio field and the number of chip select will automatically increased.

Signed-off-by Jean-Christophe PLAGNIOL-VILLARD <[email protected]>

Cc devicetree-discuss at lists.ozlabs.org

Cc spi-devel-general at lists.sourceforge.net

---
 Documentation/devicetree/bindings/spi/spi-bus.txt |    6 ++
 drivers/spi/spi.c                                 |   55 +++++++++++++++++++-
 include/linux/spi/spi.h                           |    3 +
 3 files changed, 61 insertions(+), 3 deletions(-)

Documentation/devicetree/bindings/spi/spi-bus.txt +6 -0

index e782add..c253379 100644
@@ -12,6 +12,7 @@ The SPI master node requires the following properties:
 - #size-cells     - should be zero.
 - compatible      - name of SPI bus controller following generic names
     		recommended practice.
+- cs-gpios	  - (optional) gpios chip select.
 No other properties are required in the SPI bus node.  It is assumed
 that a driver for an SPI bus device will understand that it is an SPI bus.
 However, the binding does not attempt to define the specific method for

@@ -21,6 +22,8 @@ assumption that board specific platform code will be used to manage
 chip selects.  Individual drivers can define additional properties to
 support describing the chip select layout.
 
+If cs-gpios is used the number of chip select will automatically increased.
+
 SPI slave nodes must be children of the SPI master node and can
 contain the following properties.
 - reg             - (required) chip select address of device.

@@ -34,6 +37,9 @@ contain the following properties.
 - spi-cs-high     - (optional) Empty property indicating device requires
     		chip select active high
 
+If a gpio chipselect is used for the SPI slave the gpio number will be passed
+via the controller_data
+
 SPI example for an MPC5200 SPI bus:
 	spi at f00 {
 		#address-cells = <1>;

drivers/spi/spi.c +52 -3

97 lines not on this page · raw message

include/linux/spi/spi.h +3 -0

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)

Signature · 1 line

1.7.2.5

Richard Genoud

author

2012-08-14 13:49 UTC #3 lore raw

PATCH 3/23 spi/atmel_spi: trivial: change some comments

In reply to Richard Genoud

From: Nicolas Ferre <[email protected]>

To be accurate with introduction of dmaengine enabled driver.

Signed-off-by Nicolas Ferre <[email protected]>

Signed-off-by Uwe Kleine-K?nig <[email protected]>

---
 drivers/spi/spi-atmel.c |    3 ++-
 1 files changed, 2 insertions(+), 1 deletions(-)

drivers/spi/spi-atmel.c +2 -1

index 16d6a83..7c41b07 100644
@@ -201,6 +201,7 @@ struct atmel_spi {
 	struct spi_transfer	*next_transfer;
 	unsigned long		next_remaining_bytes;
 
+	/* scratch buffer */
 	void			*buffer;
 	dma_addr_t		buffer_dma;
 };

@@ -364,7 +365,7 @@ 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,

Signature · 1 line

1.7.2.5

Richard Genoud

author

2012-08-14 13:49 UTC #4 lore raw

PATCH 4/23 spi/atmel_spi: add physical base address

In reply to Richard Genoud

From: Nicolas Ferre <[email protected]>

Needed for future use with dmaengine enabled driver.

Signed-off-by Nicolas Ferre <[email protected]>

Signed-off-by Uwe Kleine-K?nig <[email protected]>

---
 drivers/spi/spi-atmel.c |    2 ++
 1 files changed, 2 insertions(+), 0 deletions(-)

drivers/spi/spi-atmel.c +2 -0

index 7c41b07..a83f470 100644
@@ -188,6 +188,7 @@
 struct atmel_spi {
 	spinlock_t		lock;
 
+	resource_size_t		phybase;
 	void __iomem		*regs;
 	int			irq;
 	struct clk		*clk;

@@ -962,6 +963,7 @@ 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;

Signature · 1 line

1.7.2.5

Richard Genoud

author

2012-08-14 13:49 UTC #5 lore raw

PATCH 5/23 spi/atmel_spi: call unmapping on transfers buffers

In reply to Richard Genoud

From: Nicolas Ferre <[email protected]>

Signed-off-by Nicolas Ferre <[email protected]>

Signed-off-by Uwe Kleine-K?nig <[email protected]>

---
 drivers/spi/spi-atmel.c |    8 +++++---
 1 files changed, 5 insertions(+), 3 deletions(-)

drivers/spi/spi-atmel.c +5 -3

index a83f470..39285ef 100644
@@ -1011,6 +1011,7 @@ 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);

@@ -1022,9 +1023,10 @@ 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);
 	}

Signature · 1 line

1.7.2.5

Richard Genoud

author

2012-08-14 13:49 UTC #6 lore raw

PATCH 6/23 spi/atmel_spi: status information passed through controller data

In reply to Richard Genoud

From: Nicolas Ferre <[email protected]>

The status of transfer is stored in controller data structure so that it can be used not only by atmel_spi_msg_done() function. This will be useful for upcoming dmaengine enabled driver.

Signed-off-by Nicolas Ferre <[email protected]>

Signed-off-by Uwe Kleine-K?nig <[email protected]>

---
 drivers/spi/spi-atmel.c |   13 ++++++++-----
 1 files changed, 8 insertions(+), 5 deletions(-)

drivers/spi/spi-atmel.c +8 -5

index 39285ef..49a7f4f 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;
 
 	/* scratch buffer */
 	void			*buffer;

@@ -545,15 +546,15 @@ 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",

@@ -565,6 +566,7 @@ 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)

@@ -642,7 +644,8 @@ 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;

@@ -660,7 +663,7 @@ 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) {

Signature · 1 line

1.7.2.5

Richard Genoud

author

2012-08-14 13:49 UTC #7 lore raw

PATCH 7/23 spi/atmel_spi: add flag to controller data for lock operations

In reply to Richard Genoud

From: Nicolas Ferre <[email protected]>

Will allow to drop the lock during DMA operations.

Signed-off-by Nicolas Ferre <[email protected]>

Signed-off-by Uwe Kleine-K?nig <[email protected]>

---
 drivers/spi/spi-atmel.c |   31 +++++++++++++++++++------------
 1 files changed, 19 insertions(+), 12 deletions(-)

drivers/spi/spi-atmel.c +19 ~~-12~~

index 49a7f4f..70f7bfa 100644
@@ -187,6 +187,7 @@
  */
 struct atmel_spi {
 	spinlock_t		lock;
+	unsigned long		flags;
 
 	resource_size_t		phybase;
 	void __iomem		*regs;

@@ -324,6 +325,16 @@ 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)
 {

@@ -560,9 +571,9 @@ 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;

@@ -789,13 +800,11 @@ 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;

@@ -814,7 +823,6 @@ 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;

@@ -879,11 +887,11 @@ 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;
 }

@@ -893,17 +901,16 @@ 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);

Signature · 1 line

1.7.2.5

Richard Genoud

author

2012-08-14 13:49 UTC #8 lore raw

PATCH 8/23 spi/atmel: add DT support

In reply to Richard Genoud

From: Jean-Christophe PLAGNIOL-VILLARD <[email protected]>

The atmel_spi use only gpio for chip select.

Signed-off-by Jean-Christophe PLAGNIOL-VILLARD <[email protected]>

Cc devicetree-discuss at lists.ozlabs.org

Cc spi-devel-general at lists.sourceforge.net

---
 .../devicetree/bindings/spi/spi_atmel.txt          |    6 +++++
 drivers/spi/spi-atmel.c                            |   21 ++++++++++++++++---
 2 files changed, 23 insertions(+), 4 deletions(-)
 create mode 100644 Documentation/devicetree/bindings/spi/spi_atmel.txt

Documentation/devicetree/bindings/spi/spi_atmel.txt new file +6 -0

new file mode 100644
index 0000000..7ec3d8d
@@ -0,0 +1,6 @@
+Atmel SPI device
+
+Required properties:
+- compatible : should be "atmel,at91rm9200-spi".
+- reg: Address and length of the register set for the device
+- interrupts: Should contain macb interrupt

drivers/spi/spi-atmel.c +17 -4

index 70f7bfa..ad47b6d 100644
@@ -19,6 +19,7 @@
 #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>

@@ -711,7 +712,7 @@ 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);

@@ -783,7 +784,9 @@ 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);

@@ -900,7 +903,7 @@ 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;

@@ -950,7 +953,8 @@ 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;

@@ -1079,11 +1083,20 @@ 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,

Signature · 1 line

1.7.2.5

Richard Genoud

author

2012-08-14 13:49 UTC #9 lore raw

PATCH 9/23 spi/atmel_spi: add dmaengine support

In reply to Richard Genoud

From: Nicolas Ferre <[email protected]>

Signed-off-by Nicolas Ferre <[email protected]>

Signed-off-by Uwe Kleine-K?nig <[email protected]>

---
 drivers/spi/Kconfig     |    9 +
 drivers/spi/spi-atmel.c |  483 ++++++++++++++++++++++++++++++++++++++++++++++-
 2 files changed, 482 insertions(+), 10 deletions(-)

drivers/spi/Kconfig +9 -0

index 5f84b55..169592a 100644
@@ -74,6 +74,15 @@ config SPI_ATMEL
 	  This selects a driver for the Atmel SPI Controller, present on
 	  many AT32 (AVR32) and AT91 (ARM) chips.
 
+config SPI_ATMEL_DMA
+	bool "Atmel SPI DMA support"
+	depends on SPI_ATMEL && (ARCH_AT91SAM9G45 || ARCH_AT91SAM9X5) && DMA_ENGINE && EXPERIMENTAL
+	default y
+	help
+	  Say Y here if you want the Atmel SPI driver to use the DMA engine. Data transfers
+	  will be handled by the DMA controller: it will increase throughput and reduce
+	  CPU utilization.
+
 config SPI_BFIN5XX
 	tristate "SPI controller driver for ADI Blackfin5xx"
 	depends on BLACKFIN

drivers/spi/spi-atmel.c +473 ~~-10~~

599 lines not on this page · raw message

Signature · 1 line

1.7.2.5

Richard Genoud

author

2012-08-14 13:49 UTC #10 lore raw

PATCH 10/23 spi-atmel: update with dmaengine interface

In reply to Richard Genoud

commit 185ecb5f4fd43911c35956d4cc7d94a1da30417f changed dmaengine interface.
=> we have to update spi-atmel accordingly

Signed-off-by Richard Genoud <[email protected]>

---
 drivers/spi/spi-atmel.c |    6 ++++--
 1 files changed, 4 insertions(+), 2 deletions(-)

drivers/spi/spi-atmel.c +4 -2

index 9456fca..48bf3bc 100644
@@ -542,7 +542,8 @@ static int atmel_spi_next_xfer_dma(struct spi_master *master,
 					&as->dma.sgrx,
 					1,
 					DMA_FROM_DEVICE,
-					DMA_PREP_INTERRUPT | DMA_CTRL_ACK);
+					DMA_PREP_INTERRUPT | DMA_CTRL_ACK,
+					NULL);
 	if (!rxdesc)
 		goto err_dma;

@@ -550,7 +551,8 @@ static int atmel_spi_next_xfer_dma(struct spi_master *master,
 					&as->dma.sgtx,
 					1,
 					DMA_TO_DEVICE,
-					DMA_PREP_INTERRUPT | DMA_CTRL_ACK);
+					DMA_PREP_INTERRUPT | DMA_CTRL_ACK,
+					NULL);
 	if (!txdesc)
 		goto err_dma;

Signature · 1 line

1.7.2.5

Richard Genoud

author

2012-08-14 13:49 UTC #11 lore raw

PATCH 11/23 spi-atmel: fix __init/__devinit sections mismatch

In reply to Richard Genoud

Signed-off-by Richard Genoud <[email protected]>

---
 drivers/spi/spi-atmel.c |    6 +++---
 1 files changed, 3 insertions(+), 3 deletions(-)

drivers/spi/spi-atmel.c +3 -3

index 48bf3bc..d90e8fb 100644
@@ -383,7 +383,7 @@ static inline int atmel_spi_xfer_can_be_chained(struct spi_transfer *xfer)
 }
 
 #if defined(CONFIG_SPI_ATMEL_DMA)
-static bool __init filter(struct dma_chan *chan, void *slave)
+static bool filter(struct dma_chan *chan, void *slave)
 {
 	struct	at_dma_slave		*sl = slave;

@@ -395,7 +395,7 @@ static bool __init filter(struct dma_chan *chan, void *slave)
 	}
 }
 
-static int __init atmel_spi_configure_dma(struct spi_master *master)
+static int __devinit atmel_spi_configure_dma(struct spi_master *master)
 {
 	struct atmel_spi	*as = spi_master_get_devdata(master);
 	struct device		*controller = master->dev.parent;

@@ -628,7 +628,7 @@ static void atmel_spi_disable_dma_irq(struct atmel_spi *as)
 	spi_writel(as, PTCR, SPI_BIT(RXTDIS) | SPI_BIT(TXTDIS));
 }
 
-static int __init atmel_spi_configure_dma(struct spi_master *master)
+static int __devinit atmel_spi_configure_dma(struct spi_master *master)
 {
 	struct atmel_spi	*as = spi_master_get_devdata(master);

Signature · 1 line

1.7.2.5

Richard Genoud

author

2012-08-14 13:49 UTC #12 lore raw

PATCH 12/23 spi-atmel: Fix spi-atmel driver to adapt to slave_config changes

In reply to Richard Genoud

This is the following of the patch e2b35f3dbfc080f15b72834d08f04f0269dbe9be

Signed-off-by Richard Genoud <[email protected]>

---
 drivers/spi/spi-atmel.c |   45 ++++++++++++++++++++++++++++++++++++---------
 1 files changed, 36 insertions(+), 9 deletions(-)

drivers/spi/spi-atmel.c +36 -9

index d90e8fb..28a00ab 100644
@@ -399,17 +399,24 @@ static int __devinit atmel_spi_configure_dma(struct spi_master *master)
 {
 	struct atmel_spi	*as = spi_master_get_devdata(master);
 	struct device		*controller = master->dev.parent;
+	struct dma_slave_config slave_config;
 	struct at_dma_slave	*sdata;
+	int err;
 
 	sdata = controller->platform_data;
 
+	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);

@@ -417,14 +424,28 @@ static int __devinit atmel_spi_configure_dma(struct spi_master *master)
 		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;
+		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 "

@@ -433,6 +454,12 @@ static int __devinit atmel_spi_configure_dma(struct spi_master *master)
 				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)

Signature · 1 line

1.7.2.5

Richard Genoud

author

2012-08-14 13:49 UTC #13 lore raw

PATCH 13/23 AT91 DMA OF support

In reply to Richard Genoud

Signed-off-by Richard Genoud <[email protected]>

---
 arch/arm/mach-at91/include/mach/at_hdmac.h |    1 +
 drivers/of/Kconfig                         |    4 +
 drivers/of/Makefile                        |    1 +
 drivers/of/of_atmel.c                      |  112 ++++++++++++++++++++++++++++
 include/linux/of_atmel.h                   |   29 +++++++
 5 files changed, 147 insertions(+), 0 deletions(-)
 create mode 100644 drivers/of/of_atmel.c
 create mode 100644 include/linux/of_atmel.h

arch/arm/mach-at91/include/mach/at_hdmac.h +1 -0

index cab0997..35667bd 100644
@@ -52,6 +52,7 @@ struct at_dma_slave {
 #define		ATC_LOCK_IF_L_CHUNK	(0x0 << 22)
 #define		ATC_LOCK_IF_L_BUFFER	(0x1 << 22)
 #define	ATC_AHB_PROT_MASK	(0x7 << 24)	/* AHB Protection */
+#define	ATC_AHB_PROT(h)		(((h) <<  24) && ATC_AHB_PROT_MASK)
 #define	ATC_FIFOCFG_MASK	(0x3 << 28)	/* FIFO Request Configuration */
 #define		ATC_FIFOCFG_LARGESTBURST	(0x0 << 28)
 #define		ATC_FIFOCFG_HALFFIFO		(0x1 << 28)

drivers/of/Kconfig +4 -0

index dfba3e6..4667960 100644
@@ -83,4 +83,8 @@ config OF_MTD
 	depends on MTD
 	def_bool y
 
+config OF_ATMEL
+	depends on AT_HDMAC
+	def_bool y
+
 endmenu # OF

drivers/of/Makefile +1 -0

index e027f44..897de3b 100644
@@ -11,3 +11,4 @@ obj-$(CONFIG_OF_MDIO)	+= of_mdio.o
 obj-$(CONFIG_OF_PCI)	+= of_pci.o
 obj-$(CONFIG_OF_PCI_IRQ)  += of_pci_irq.o
 obj-$(CONFIG_OF_MTD)	+= of_mtd.o
+obj-$(CONFIG_OF_ATMEL)	+= of_atmel.o

drivers/of/of_atmel.c new file +112 -0

112 lines not on this page · raw message

include/linux/of_atmel.h new file +29 -0

new file mode 100644
index 0000000..256f336
@@ -0,0 +1,29 @@
+#ifndef _LINUX_OF_ATMEL_H
+#define _LINUX_OF_ATMEL_H
+/*
+ * Copyright (C) 2012 Richard Genoud, Paratronic S.A.
+ *                    <[email protected]>
+ *
+ * Atmel specifics OF helpers
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ *
+ */
+#include <linux/of.h>
+#include <linux/device.h>
+#include <mach/at_hdmac.h>
+
+#ifdef CONFIG_OF_ATMEL
+extern struct at_dma_slave *of_create_at_dma_slave(struct device_node *np,
+						   struct bus_type *bus);
+#else
+static struct at_dma_slave *of_create_at_dma_slave(struct device_node *np,
+						   struct bus_type *bus)
+{
+	return NULL;
+}
+#endif /* CONFIG_OF_ATMEL */
+
+#endif	/* _LINUX_OF_ATMEL_H */

Signature · 1 line

1.7.2.5

Nicolas Ferre

2012-08-14 14:47 UTC #14 lore raw

PATCH 13/23 AT91 DMA OF support

Hi Richard,

I am not sure that this one will go further: there is an initiative to build a pretty generic DMA OF support... It takes time but it is progressing...

Anyway, there may have some bits to integrate in at_hdmac.c driver: I will have a closer look later.

Thanks,

On 08/14/2012 03:49 PM, Richard Genoud :

Quoted 199 lines

Signed-off-by Richard Genoud <[email protected]>

---
 arch/arm/mach-at91/include/mach/at_hdmac.h |    1 +
 drivers/of/Kconfig                         |    4 +
 drivers/of/Makefile                        |    1 +
 drivers/of/of_atmel.c                      |  112 ++++++++++++++++++++++++++++
 include/linux/of_atmel.h                   |   29 +++++++
 5 files changed, 147 insertions(+), 0 deletions(-)
 create mode 100644 drivers/of/of_atmel.c
 create mode 100644 include/linux/of_atmel.h

arch/arm/mach-at91/include/mach/at_hdmac.h +1 -0

index cab0997..35667bd 100644
@@ -52,6 +52,7 @@ struct at_dma_slave {
 #define		ATC_LOCK_IF_L_CHUNK	(0x0 << 22)
 #define		ATC_LOCK_IF_L_BUFFER	(0x1 << 22)
 #define	ATC_AHB_PROT_MASK	(0x7 << 24)	/* AHB Protection */
+#define	ATC_AHB_PROT(h)		(((h) <<  24) && ATC_AHB_PROT_MASK)
 #define	ATC_FIFOCFG_MASK	(0x3 << 28)	/* FIFO Request Configuration */
 #define		ATC_FIFOCFG_LARGESTBURST	(0x0 << 28)
 #define		ATC_FIFOCFG_HALFFIFO		(0x1 << 28)

drivers/of/Kconfig +4 -0

index dfba3e6..4667960 100644
@@ -83,4 +83,8 @@ config OF_MTD
 	depends on MTD
 	def_bool y
 
+config OF_ATMEL
+	depends on AT_HDMAC
+	def_bool y
+
 endmenu # OF

drivers/of/Makefile +1 -0

index e027f44..897de3b 100644
@@ -11,3 +11,4 @@ obj-$(CONFIG_OF_MDIO)	+= of_mdio.o
 obj-$(CONFIG_OF_PCI)	+= of_pci.o
 obj-$(CONFIG_OF_PCI_IRQ)  += of_pci_irq.o
 obj-$(CONFIG_OF_MTD)	+= of_mtd.o
+obj-$(CONFIG_OF_ATMEL)	+= of_atmel.o

drivers/of/of_atmel.c new file +112 -0

new file mode 100644
index 0000000..bcd3c54a
@@ -0,0 +1,112 @@
+/*
+ * Copyright (C) 2012 Richard Genoud, Paratronic S.A.
+ *                    <[email protected]>
+ *
+ * Atmel specifics OF helpers
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ *
+ */
+#include <linux/kernel.h>
+#include <linux/export.h>
+#include <linux/device.h>
+#include <linux/slab.h>
+#include <linux/printk.h>
+#include <linux/byteorder/generic.h>
+#include <mach/at_hdmac.h>
+#include <linux/of.h>
+
+static int of_dev_node_match(struct device *dev, void *data)
+{
+	return dev->of_node == data;
+}
+
+/**
+ * of_create_at_dma_slave - Alloc and initialize the struct at_dma_slave
+ * @np: pointer to node to create the struct for
+ * @bus: pointer to the bus type of the DMA device.
+ *
+ * Returns pointer to created DMA slave structure, or NULL in case of error.
+ */
+struct at_dma_slave *of_create_at_dma_slave(struct device_node *np,
+					    struct bus_type *bus)
+{
+	struct at_dma_slave *atslave;
+	struct device_node *n;
+	struct property *prop;
+	const __be32 *dma_list;
+	phandle phandle;
+	int size;
+	int err = -1;
+	u32 val;
+
+	atslave = kzalloc(sizeof(struct at_dma_slave), GFP_KERNEL);
+	if (unlikely(!atslave)) {
+		pr_err("cannot allocate memory\n");
+		goto error;
+	}
+
+	atslave->cfg |= of_property_read_bool(np, "atc_src_rep") ? ATC_SRC_REP : 0;
+	atslave->cfg |= of_property_read_bool(np, "atc_dst_rep") ? ATC_DST_REP : 0;
+	atslave->cfg |= of_property_read_bool(np, "atc_src_h2sel_hw") ? ATC_SRC_H2SEL_HW : 0;
+	atslave->cfg |= of_property_read_bool(np, "atc_dst_h2sel_hw") ? ATC_DST_H2SEL_HW : 0;
+	atslave->cfg |= of_property_read_bool(np, "atc_sod") ? ATC_SOD : 0;
+	atslave->cfg |= of_property_read_bool(np, "atc_lock_if") ? ATC_LOCK_IF : 0;
+	atslave->cfg |= of_property_read_bool(np, "atc_lock_b") ? ATC_LOCK_B : 0;
+	atslave->cfg |= of_property_read_bool(np, "atc_lock_if_l") ? ATC_LOCK_IF_L : 0;
+	atslave->cfg |= of_property_read_bool(np, "atc_lock_if_l_buffer") ? ATC_LOCK_IF_L_BUFFER : 0;
+	if (of_property_read_bool(np, "atc_fifocfg_halffifo"))
+		atslave->cfg = (atslave->cfg & ~ATC_FIFOCFG_MASK) | ATC_FIFOCFG_HALFFIFO;
+	if (of_property_read_bool(np, "atc_fifocfg_enoughspace"))
+		atslave->cfg = (atslave->cfg & ~ATC_FIFOCFG_MASK) | ATC_FIFOCFG_ENOUGHSPACE;
+	if (of_property_read_bool(np, "atc_fifocfg_largestburst"))
+		atslave->cfg = (atslave->cfg & ~ATC_FIFOCFG_MASK) | ATC_FIFOCFG_LARGESTBURST;
+	if (!of_property_read_u32(np, "atc_ahb_prot", &val))
+		atslave->cfg |= ATC_AHB_PROT(val);
+	if (!of_property_read_u32(np, "atc_src_per", &val))
+		atslave->cfg |= ATC_SRC_PER(val);
+	if (!of_property_read_u32(np, "atc_dst_per", &val))
+		atslave->cfg |= ATC_DST_PER(val);
+
+	/* we need to associate the wanted dma struct device to
+	 * atslave->dma_dev. This is needed when searching for
+	 * a free dma channel
+	 */
+	prop = of_find_property(np, "dma", &size);
+	if (!prop) {
+		pr_err("can't find the dma child node\n");
+		goto error;
+	}
+	dma_list = prop->value;
+	size /= sizeof(*dma_list);
+	if (size != 1) {
+		pr_err("only one dma handle is supported\n");
+		goto error;
+	}
+
+	phandle = be32_to_cpup(dma_list);
+	n = of_find_node_by_phandle(phandle);
+	if (!n) {
+		pr_err("unable to find dma device: invalid phandle %s\n",
+		       prop->name);
+		goto error;
+	}
+
+	atslave->dma_dev = bus_find_device(bus, NULL, n, of_dev_node_match);
+	if (!atslave->dma_dev) {
+		pr_err("can't find struct device for DMA %s\n", prop->name);
+		goto error;
+	}
+
+	pr_debug("DMA cfg register=0x%x DMA device path=%s\n",
+		 atslave->cfg, prop->name);
+	err = 0;
+error:
+	if (err)
+		kfree(atslave);
+	return atslave;
+}
+EXPORT_SYMBOL_GPL(of_create_at_dma_slave);
+

include/linux/of_atmel.h new file +29 -0

new file mode 100644
index 0000000..256f336
@@ -0,0 +1,29 @@
+#ifndef _LINUX_OF_ATMEL_H
+#define _LINUX_OF_ATMEL_H
+/*
+ * Copyright (C) 2012 Richard Genoud, Paratronic S.A.
+ *                    <[email protected]>
+ *
+ * Atmel specifics OF helpers
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ *
+ */
+#include <linux/of.h>
+#include <linux/device.h>
+#include <mach/at_hdmac.h>
+
+#ifdef CONFIG_OF_ATMEL
+extern struct at_dma_slave *of_create_at_dma_slave(struct device_node *np,
+						   struct bus_type *bus);
+#else
+static struct at_dma_slave *of_create_at_dma_slave(struct device_node *np,
+						   struct bus_type *bus)
+{
+	return NULL;
+}
+#endif /* CONFIG_OF_ATMEL */
+
+#endif	/* _LINUX_OF_ATMEL_H */

Signature · 1 line

Nicolas Ferre

Richard Genoud

author

2012-08-14 13:49 UTC #15 lore raw

PATCH 14/23 add at91sam9x5 Kconfig ARCH/SOC link

In reply to Richard Genoud

---
 arch/arm/mach-at91/Kconfig |    4 ++++
 1 files changed, 4 insertions(+), 0 deletions(-)

arch/arm/mach-at91/Kconfig +4 -0

index c8050b1..c6eb877 100644
@@ -141,6 +141,10 @@ config ARCH_AT91SAM9G45
 	bool "AT91SAM9G45"
 	select SOC_AT91SAM9G45
 
+config ARCH_AT91SAM9X5
+	bool "AT91SAM9x5"
+	select SOC_AT91SAM9X5
+
 config ARCH_AT91X40
 	bool "AT91x40"
 	depends on !MMU

Signature · 1 line

1.7.2.5

Richard Genoud

author

2012-08-14 13:49 UTC #16 lore raw

PATCH 15/23 spi-atmel: add DMA OF support

In reply to Richard Genoud

Signed-off-by Richard Genoud <[email protected]>

---
 drivers/of/of_atmel.c   |    1 -
 drivers/spi/spi-atmel.c |   22 ++++++++++++++++++++++
 2 files changed, 22 insertions(+), 1 deletions(-)

drivers/of/of_atmel.c +0 -1

index bcd3c54a..a59b9ced 100644
@@ -55,7 +55,6 @@ struct at_dma_slave *of_create_at_dma_slave(struct device_node *np,
 	atslave->cfg |= of_property_read_bool(np, "atc_sod") ? ATC_SOD : 0;
 	atslave->cfg |= of_property_read_bool(np, "atc_lock_if") ? ATC_LOCK_IF : 0;
 	atslave->cfg |= of_property_read_bool(np, "atc_lock_b") ? ATC_LOCK_B : 0;
-	atslave->cfg |= of_property_read_bool(np, "atc_lock_if_l") ? ATC_LOCK_IF_L : 0;
 	atslave->cfg |= of_property_read_bool(np, "atc_lock_if_l_buffer") ? ATC_LOCK_IF_L_BUFFER : 0;
 	if (of_property_read_bool(np, "atc_fifocfg_halffifo"))
 		atslave->cfg = (atslave->cfg & ~ATC_FIFOCFG_MASK) | ATC_FIFOCFG_HALFFIFO;

drivers/spi/spi-atmel.c +22 -0

index 28a00ab..d629422 100644
@@ -21,6 +21,7 @@
 #include <linux/spi/spi.h>
 #include <linux/slab.h>
 #include <linux/of.h>
+#include <linux/of_atmel.h>
 
 #include <asm/io.h>
 #include <mach/board.h>

@@ -1411,6 +1412,8 @@ static int __devinit atmel_spi_probe(struct platform_device *pdev)
 	int			ret;
 	struct spi_master	*master;
 	struct atmel_spi	*as;
+	struct bus_type		*bus = pdev->dev.bus;
+	struct device_node	*of = pdev->dev.of_node;
 
 	regs = platform_get_resource(pdev, IORESOURCE_MEM, 0);
 	if (!regs)

@@ -1430,6 +1433,17 @@ static int __devinit atmel_spi_probe(struct platform_device *pdev)
 	if (!master)
 		goto out_free;
 
+
+	/* populate platform_data from device tree */
+	if (of) {
+		pdev->dev.platform_data = of_create_at_dma_slave(of, bus);
+		if (!pdev->dev.platform_data) {
+			dev_err(&pdev->dev, "could not get DMA\n");
+			ret = -EINVAL;
+			goto out_free;
+		}
+	}
+
 	/* the spi->mode bits understood by this driver: */
 	master->mode_bits = SPI_CPOL | SPI_CPHA | SPI_CS_HIGH;

@@ -1504,6 +1518,10 @@ out_free_buffer:
 	dma_free_coherent(&pdev->dev, BUFFER_SIZE, as->buffer,
 			as->buffer_dma);
 out_free:
+	if (of) {
+		kfree((struct at_dma_slave*)(pdev->dev.platform_data));
+		pdev->dev.platform_data = NULL;
+	}
 	clk_put(clk);
 	spi_master_put(master);
 	return ret;

@@ -1546,6 +1564,10 @@ static int __devexit atmel_spi_remove(struct platform_device *pdev)
 	iounmap(as->regs);
 
 	spi_unregister_master(master);
+	if (pdev->dev.of_node) {
+		kfree((struct at_dma_slave*)(pdev->dev.platform_data));
+		pdev->dev.platform_data = NULL;
+	}
 
 	return 0;
 }

Signature · 1 line

1.7.2.5

Richard Genoud

author

2012-08-14 13:49 UTC #17 lore raw

PATCH 16/23 SPI: array out of bound => no CS

In reply to Richard Genoud

---
 drivers/spi/spi.c |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)

drivers/spi/spi.c +1 -1

index 3fb5b6d..74e6577 100644
@@ -976,8 +976,8 @@ static int of_spi_register_master(struct spi_master *master)
 		return -ENOMEM;
 
 	memset(cs, -EINVAL, master->num_chipselect);
-	master->num_chipselect += nb;
 	cs += master->num_chipselect;
+	master->num_chipselect += nb;
 
 	for (i = 0; i < nb; i++)
 		cs[i] = of_get_named_gpio(np, "cs-gpios", i);

Signature · 1 line

1.7.2.5

Richard Genoud

author

2012-08-14 13:49 UTC #18 lore raw

PATCH 17/23 atmel-spi && DMA: OOPS if buffer > 4400 bytes

In reply to Richard Genoud

this is just a quick and suboptimal workaround

---
 drivers/spi/spi-atmel.c |    3 ++-
 1 files changed, 2 insertions(+), 1 deletions(-)

drivers/spi/spi-atmel.c +2 -1

index d629422..0899da2 100644
@@ -638,7 +638,8 @@ static void atmel_spi_next_xfer(struct spi_master *master,
 
 	as->current_transfer = xfer;
 
-	if (atmel_spi_use_dma(xfer)) {
+	/* quick (and *really* not optimal) workaround for DMA BUG */
+	if (atmel_spi_use_dma(xfer) && (xfer->len < BUFFER_SIZE)) {
 		if (!atmel_spi_next_xfer_dma(master, xfer))
 			return;
 		else

Signature · 1 line

1.7.2.5

Richard Genoud

author

2012-08-14 13:49 UTC #19 lore raw

PATCH 18/23 sam9x5: declare SPI clocks

In reply to Richard Genoud

Signed-off-by Richard Genoud <[email protected]>

---
 arch/arm/mach-at91/at91sam9x5.c |    3 +++
 1 files changed, 3 insertions(+), 0 deletions(-)

arch/arm/mach-at91/at91sam9x5.c +3 -0

index bbd2f8e..e213e2d 100644
@@ -16,6 +16,7 @@
 #include <mach/at91_pmc.h>
 #include <mach/cpu.h>
 #include <mach/board.h>
+#include <mach/gpio.h>
 
 #include "soc.h"
 #include "generic.h"

@@ -231,6 +232,8 @@ 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_ID("pioA", &pioAB_clk),
 	CLKDEV_CON_ID("pioB", &pioAB_clk),
 	CLKDEV_CON_ID("pioC", &pioCD_clk),

Signature · 1 line

1.7.2.5

Richard Genoud

author

2012-08-14 13:49 UTC #20 lore raw

PATCH 19/23 spi-atmel: add sam9x5 SPI in device tree

In reply to Richard Genoud

Signed-off-by Richard Genoud <[email protected]>

---
 arch/arm/boot/dts/at91sam9x5.dtsi |   30 ++++++++++++++++++++++++++++++
 1 files changed, 30 insertions(+), 0 deletions(-)

arch/arm/boot/dts/at91sam9x5.dtsi +30 -0

index e58965d6..8cb8b0d 100644
@@ -344,6 +344,36 @@
 					trigger-value = <0x6>;
 				};
 			};
+
+			spi0: spi at f0000000 {
+				#address-cells = <1>;
+				#size-cells = <0>;
+				compatible = "atmel,at91rm9200-spi";
+				reg = <0xf0000000 0x100>;
+				interrupts = <13 4 3>;
+				cs-gpios = <&pioA 14 0
+					    &pioA 7 0 /* conflicts with TXD2 */
+					    &pioA 1 0 /* conflicts with RXD0 */
+					    &pioB 3 0 /* conflicts with ERXDV */
+					   >;
+				dma-mask = <0xffffffff>;
+				status = "disabled";
+			};
+
+			spi1: spi at f0004000 {
+				#address-cells = <1>;
+				#size-cells = <0>;
+				compatible = "atmel,at91rm9200-spi";
+				reg = <0xf0004000 0x100>;
+				interrupts = <14 4 3>;
+				cs-gpios = <&pioA 8 0 /* conflitcs with RXD2 */
+					    &pioA 0 0 /* conflitcs with TXD0 */
+					    &pioA 31 0 /* conflitcs with TWCK0 */
+					    &pioA 30 0 /* conflitcs with TWD0 */
+					   >;
+				dma-mask = <0xffffffff>;
+				status = "disabled";
+			};
 		};
 
 		nand0: nand at 40000000 {

Signature · 1 line

1.7.2.5

Richard Genoud

author

2012-08-14 13:49 UTC #21 lore raw

PATCH 20/23 spi-atmel: add dma support in sam9x5 device tree

In reply to Richard Genoud

Signed-off-by Richard Genoud <[email protected]>

---
 arch/arm/boot/dts/at91sam9x5.dtsi |   12 ++++++++++++
 1 files changed, 12 insertions(+), 0 deletions(-)

arch/arm/boot/dts/at91sam9x5.dtsi +12 -0

index 8cb8b0d..ef09816 100644
@@ -357,6 +357,12 @@
 					    &pioB 3 0 /* conflicts with ERXDV */
 					   >;
 				dma-mask = <0xffffffff>;
+				dma = <&dma0>;
+				atc_src_per = <2>; /* AT_DMA_ID_SPI0_RX */
+				atc_dst_per = <1>; /* AT_DMA_ID_SPI0_TX */
+				atc_fifocfg_halffifo;
+				atc_src_h2sel_hw;
+				atc_dst_h2sel_hw;
 				status = "disabled";
 			};

@@ -372,6 +378,12 @@
 					    &pioA 30 0 /* conflitcs with TWD0 */
 					   >;
 				dma-mask = <0xffffffff>;
+				dma = <&dma0>;
+				atc_src_per = <4>; /* AT_DMA_ID_SPI1_RX */
+				atc_dst_per = <3>; /* AT_DMA_ID_SPI1_TX */
+				atc_fifocfg_halffifo;
+				atc_src_h2sel_hw;
+				atc_dst_h2sel_hw;
 				status = "disabled";
 			};
 		};

Signature · 1 line

1.7.2.5

Richard Genoud

author

2012-08-14 13:49 UTC #22 lore raw

PATCH 21/23 spi-atmel OF: complete documentation

In reply to Richard Genoud

Signed-off-by Richard Genoud <[email protected]>

---
 .../devicetree/bindings/spi/spi_atmel.txt          |   15 +++++++++++++++
 1 files changed, 15 insertions(+), 0 deletions(-)

Documentation/devicetree/bindings/spi/spi_atmel.txt +15 -0

index 7ec3d8d..6d2e5e1 100644
@@ -4,3 +4,18 @@ Required properties:
 - compatible : should be "atmel,at91rm9200-spi".
 - reg: Address and length of the register set for the device
 - interrupts: Should contain macb interrupt
+- cs-gpio: Should contain the GPIOs used for chipselect.
+
+spi0: spi at f0000000 {
+	#address-cells = <1>;
+	#size-cells = <0>;
+	compatible = "atmel,at91rm9200-spi";
+	reg = <0xf0000000 0x100>;
+	interrupts = <13 4>;
+	cs-gpios = <&pioA 14 0
+		    &pioA 7 0 /* conflicts with TXD2 */
+		    &pioA 1 0 /* conflicts with RXD0 */
+		    &pioB 3 0 /* conflicts with ERXDV */
+		   >;
+	status = "disabled";
+};

Signature · 1 line

1.7.2.5

Richard Genoud

author

2012-08-14 13:49 UTC #23 lore raw

PATCH 22/23 spi-atmel: complete DMA slave OF documentation

In reply to Richard Genoud

Signed-off-by Richard Genoud <[email protected]>

---
 .../devicetree/bindings/spi/spi_atmel.txt          |   25 ++++++++++++++++++++
 1 files changed, 25 insertions(+), 0 deletions(-)

Documentation/devicetree/bindings/spi/spi_atmel.txt +25 -0

index 6d2e5e1..b5d8082 100644
@@ -5,6 +5,24 @@ Required properties:
 - reg: Address and length of the register set for the device
 - interrupts: Should contain macb interrupt
 - cs-gpio: Should contain the GPIOs used for chipselect.
+- dma-mask: device coherent dma mask.
+/* dma transfer configuration. Cf DMA Channel configuration register */
+- dma: handle to the dma controller that should be used.
+- atc_src_per: source channel for DMA operation.
+- atc_dst_per: destination channel for DMA operation.
+- atc_src_h2sel_hw: Hardware handshaking interface is used to trigger a
+transfer request.
+- atc_dst_h2sel_hw: Hardware handshaking interface is used to trigger a
+transfer request.
+- atc_sod: STOP ON DONE activated.
+- atc_lock_if: Interface Lock capability is enabled.
+- atc_lock_b: AHB Bus Locking capability is enabled.
+- atc_lock_if_l_buffer: The Master Interface Arbiter is locked by the channel x for a buffer transfer.
+- atc_fifocfg_halffifo: When half FIFO size is available/filled, a source/destination request is serviced.
+- atc_fifocfg_enoughspace: When there is enough space/data available to perform a single AHB access, then the request is serviced.
+- atc_fifocfg_largestburst: The largest defined length AHB burst is performed on the destination AHB interface.
+- atc_ahb_prot: AHB Protection value. (cf release manual)
+
 
 spi0: spi at f0000000 {
 	#address-cells = <1>;

@@ -17,5 +35,12 @@ spi0: spi at f0000000 {
 		    &pioA 1 0 /* conflicts with RXD0 */
 		    &pioB 3 0 /* conflicts with ERXDV */
 		   >;
+	dma-mask = <0xffffffff>;
+	dma = <&dma0>;
+	atc_src_per = <2>; /* AT_DMA_ID_SPI0_RX */
+	atc_dst_per = <1>; /* AT_DMA_ID_SPI0_TX */
+	atc_fifocfg_halffifo;
+	atc_src_h2sel_hw;
+	atc_dst_h2sel_hw;
 	status = "disabled";
 };

Signature · 1 line

1.7.2.5

Richard Genoud

author

2012-08-14 13:49 UTC #24 lore raw

PATCH 23/23 sam9x5ek DTS: enable SPI dataflash

In reply to Richard Genoud

Signed-off-by Richard Genoud <[email protected]>

---
 arch/arm/boot/dts/at91sam9x5ek.dtsi |   10 ++++++++++
 1 files changed, 10 insertions(+), 0 deletions(-)

arch/arm/boot/dts/at91sam9x5ek.dtsi +10 -0

index e5000a7..c99f1cc 100644
@@ -30,6 +30,16 @@
 				phy-mode = "rmii";
 				status = "okay";
 			};
+
+			spi0: spi at f0000000 {
+				status = "okay";
+				cs-gpios = <&pioA 14 0>;
+				m25p80 at 0 {
+					compatible = "atmel,at25df321a";
+					spi-max-frequency = <50000000>;
+					reg = <0>;
+				};
+			};
 		};
 
 		usb0: ohci at 00600000 {

Signature · 1 line

1.7.2.5

Richard Genoud

author

2012-08-14 14:16 UTC #25 lore raw

cover work in progress: SPI controller w/DMA SAM9X5

In reply to Richard Genoud

2012/8/14 Richard Genoud <[email protected]>:

Quoted 4 lines

Nicolas,
I said that I had some work done on sam9g35-ek, here is the spi part, based on your and Jean-Christophe work.
You have to enable CONFIG_ARCH_AT91SAM9X5 to get it work.

I forgot something:
As I didn't get pinctrl to work, on top of this patchset, I did a revert of "ARM: at91: add pinctrl support"
and I made this change :

arch/arm/mach-at91/at91sam9x5.c +8 -0

index e213e2d..bc3fdbf 100644
@@ -313,6 +313,13 @@ static void __init at91sam9x5_map_io(void)
 	at91_init_sram(0, AT91SAM9X5_SRAM_BASE, AT91SAM9X5_SRAM_SIZE);
 }
 
+
+void __init at91sam9x5_initialize(void)
+{
+	/* Register GPIO subsystem (using DT) */
+	at91_gpio_init(NULL, 0);
+}
+
 /* --------------------------------------------------------------------
  *  Interrupt initialization
  * -------------------------------------------------------------------- */

@@ -320,4 +327,5 @@ static void __init at91sam9x5_map_io(void)
 struct at91_init_soc __initdata at91sam9x5_soc = {
 	.map_io = at91sam9x5_map_io,
 	.register_clocks = at91sam9x5_register_clocks,
+	.init = at91sam9x5_initialize,
 };

I also enabled the spi controller in the DTS:

(quoted diff) +0 -0

b/arch/arm/boot/dts/at91sam9x5.dtsi
index ef09816..cc1c830 100644

arch/arm/boot/dts/at91sam9x5.dtsi +1 -1

@@ -363,7 +363,7 @@
 				atc_fifocfg_halffifo;
 				atc_src_h2sel_hw;
 				atc_dst_h2sel_hw;
-				status = "disabled";
+				status = "okay";
 			};
 
 			spi1: spi at f0004000 {

Signature · 1 line

for me, ck means con kolivas and not calvin klein... does it mean I'm a geek ?

Nicolas Ferre

2012-08-14 15:29 UTC #26 lore raw

cover work in progress: SPI controller w/DMA SAM9X5

In reply to Richard Genoud

On 08/14/2012 03:49 PM, Richard Genoud :

Quoted 11 lines

Nicolas,
I said that I had some work done on sam9g35-ek, here is the spi part, based on your and Jean-Christophe work.
You have to enable CONFIG_ARCH_AT91SAM9X5 to get it work.
The goal of this patchset is to share the work I've done, it's not ready for upstream, but it may save some time.
It's based on 3.6-rc1 + Jean-Christophe's pinctrl patchset.
Best Regards,

Richard,

Thanks a lot for updating this material: We really appreciate.

We will try to review this material together with Jean-Christophe and colleagues @ Atmel. I hope that we will go forward with this collected material in the near future.

Maybe DMA OF part will not be upstreamed as-is because of an ongoing initiative to build a generic dmaengine OF support.

Quoted 3 lines

Jean-Christophe PLAGNIOL-VILLARD (3):
of: add dma-mask binding
of_spi: add generic binding support to specify cs gpio

I know that Jean-Christophe has had difficulties to have feedback on there parts...

Quoted 10 lines

spi/atmel: add DT support
Nicolas Ferre (6):
spi/atmel_spi: trivial: change some comments spi/atmel_spi: add physical base address spi/atmel_spi: call unmapping on transfers buffers spi/atmel_spi: status information passed through controller data spi/atmel_spi: add flag to controller data for lock operations spi/atmel_spi: add dmaengine support

Quoted 2 lines

Richard Genoud (14):
spi-atmel: update with dmaengine interface

This will certainly be squashed in previous path.

Quoted 1 line

spi-atmel: fix __init/__devinit sections mismatch

This one also: but it is good to have them separated at fist.

Quoted 3 lines

spi-atmel: Fix spi-atmel driver to adapt to slave_config changes AT91 DMA OF support
add at91sam9x5 Kconfig ARCH/SOC link

Is it fixup for a bug? Jean-Christophe, a though about this one?

Quoted 7 lines

spi-atmel: add DMA OF support
[BUG] SPI: array out of bound => no CS [BUG] atmel-spi && DMA: OOPS if buffer > 4400 bytes
sam9x5: declare SPI clocks
spi-atmel: add sam9x5 SPI in device tree
spi-atmel: add dma support in sam9x5 device tree spi-atmel OF: complete documentation

Maybe documentation has to be folded in the patch that actually adds the OF support.

Quoted 24 lines

spi-atmel: complete DMA slave OF documentation
sam9x5ek DTS: enable SPI dataflash

 Documentation/devicetree/bindings/spi/spi-bus.txt  |    6 +
 .../devicetree/bindings/spi/spi_atmel.txt          |   46 ++
 arch/arm/boot/dts/at91sam9x5.dtsi                  |   42 ++
 arch/arm/boot/dts/at91sam9x5ek.dtsi                |   10 +
 arch/arm/mach-at91/Kconfig                         |    4 +
 arch/arm/mach-at91/at91sam9x5.c                    |    3 +
 arch/arm/mach-at91/include/mach/at_hdmac.h         |    1 +
 drivers/of/Kconfig                                 |    4 +
 drivers/of/Makefile                                |    1 +
 drivers/of/of_atmel.c                              |  111 ++++
 drivers/of/platform.c                              |   23 +-
 drivers/spi/Kconfig                                |    9 +
 drivers/spi/spi-atmel.c                            |  611 ++++++++++++++++++--
 drivers/spi/spi.c                                  |   55 ++-
 include/linux/of_atmel.h                           |   29 +
 include/linux/spi/spi.h                            |    3 +
 16 files changed, 914 insertions(+), 44 deletions(-)
 create mode 100644 Documentation/devicetree/bindings/spi/spi_atmel.txt
 create mode 100644 drivers/of/of_atmel.c
 create mode 100644 include/linux/of_atmel.h

Signature · 1 line

Nicolas Ferre

Jean-Christophe PLAGNIOL-VILLARD

2012-08-14 17:44 UTC #27 lore raw

cover work in progress: SPI controller w/DMA SAM9X5

Quoted 24 lines · Nicolas Ferre

On 08/14/2012 03:49 PM, Richard Genoud :
Quoted 11 lines
Nicolas,
I said that I had some work done on sam9g35-ek, here is the spi part, based on your and Jean-Christophe work.
You have to enable CONFIG_ARCH_AT91SAM9X5 to get it work.
The goal of this patchset is to share the work I've done, it's not ready for upstream, but it may save some time.
It's based on 3.6-rc1 + Jean-Christophe's pinctrl patchset.
Best Regards,
Richard,
Thanks a lot for updating this material: We really appreciate.
We will try to review this material together with Jean-Christophe and colleagues @ Atmel. I hope that we will go forward with this collected material in the near future.
Maybe DMA OF part will not be upstreamed as-is because of an ongoing initiative to build a generic dmaengine OF support.

Quoted 6 lines

Quoted 3 lines
Jean-Christophe PLAGNIOL-VILLARD (3):
of: add dma-mask binding
of_spi: add generic binding support to specify cs gpio
I know that Jean-Christophe has had difficulties to have feedback on there parts...

yes I need to re-write it as suggest by Grant

will take a look later next week on this

Quoted 18 lines

Quoted 11 lines
spi/atmel: add DT support
Nicolas Ferre (6):
spi/atmel_spi: trivial: change some comments spi/atmel_spi: add physical base address spi/atmel_spi: call unmapping on transfers buffers spi/atmel_spi: status information passed through controller data spi/atmel_spi: add flag to controller data for lock operations spi/atmel_spi: add dmaengine support
Quoted 2 lines
Richard Genoud (14):
spi-atmel: update with dmaengine interface
This will certainly be squashed in previous path.
Quoted 1 line
spi-atmel: fix __init/__devinit sections mismatch

Quoted 7 lines

This one also: but it is good to have them separated at fist.
Quoted 3 lines
spi-atmel: Fix spi-atmel driver to adapt to slave_config changes AT91 DMA OF support
add at91sam9x5 Kconfig ARCH/SOC link
Is it fixup for a bug? Jean-Christophe, a though about this one?

no the ARCH_xxx is old style can not be used on DT only soc

this patch need to be droped

Best Regards,
J.

Nicolas Ferre

2012-08-14 17:48 UTC #28 lore raw

cover work in progress: SPI controller w/DMA SAM9X5

On 08/14/2012 07:44 PM, Jean-Christophe PLAGNIOL-VILLARD :

Quoted 54 lines · Nicolas Ferre

Quoted 24 lines
On 08/14/2012 03:49 PM, Richard Genoud :
Quoted 11 lines
Nicolas,
I said that I had some work done on sam9g35-ek, here is the spi part, based on your and Jean-Christophe work.
You have to enable CONFIG_ARCH_AT91SAM9X5 to get it work.
The goal of this patchset is to share the work I've done, it's not ready for upstream, but it may save some time.
It's based on 3.6-rc1 + Jean-Christophe's pinctrl patchset.
Best Regards,
Richard,
Thanks a lot for updating this material: We really appreciate.
We will try to review this material together with Jean-Christophe and colleagues @ Atmel. I hope that we will go forward with this collected material in the near future.
Maybe DMA OF part will not be upstreamed as-is because of an ongoing initiative to build a generic dmaengine OF support.
Quoted 6 lines
Quoted 3 lines
Jean-Christophe PLAGNIOL-VILLARD (3):
of: add dma-mask binding
of_spi: add generic binding support to specify cs gpio
I know that Jean-Christophe has had difficulties to have feedback on there parts...
yes I need to re-write it as suggest by Grant
will take a look later next week on this
Quoted 20 lines
Quoted 11 lines
spi/atmel: add DT support
Nicolas Ferre (6):
spi/atmel_spi: trivial: change some comments spi/atmel_spi: add physical base address spi/atmel_spi: call unmapping on transfers buffers spi/atmel_spi: status information passed through controller data spi/atmel_spi: add flag to controller data for lock operations spi/atmel_spi: add dmaengine support
Quoted 2 lines
Richard Genoud (14):
spi-atmel: update with dmaengine interface
This will certainly be squashed in previous path.
Quoted 1 line
spi-atmel: fix __init/__devinit sections mismatch
This one also: but it is good to have them separated at fist.

Quoted 8 lines

Quoted 5 lines
Quoted 3 lines
spi-atmel: Fix spi-atmel driver to adapt to slave_config changes AT91 DMA OF support
add at91sam9x5 Kconfig ARCH/SOC link
Is it fixup for a bug? Jean-Christophe, a though about this one?
no the ARCH_xxx is old style can not be used on DT only soc
this patch need to be droped

That is what I suspected, thanks J.

Bye,

Signature · 1 line

Nicolas Ferre

Richard Genoud

author

2012-08-16 07:07 UTC #29 lore raw

cover work in progress: SPI controller w/DMA SAM9X5

2012/8/14 Nicolas Ferre <[email protected]>:

Quoted 8 lines

Quoted 6 lines
Quoted 3 lines
Quoted 1 line
add at91sam9x5 Kconfig ARCH/SOC link
Is it fixup for a bug? Jean-Christophe, a though about this one?
no the ARCH_xxx is old style can not be used on DT only soc
this patch need to be droped
That is what I suspected, thanks J.

Yes, Jean-Christophe is right. This patch needs to be dropped, and this one should be applied instead:

drivers/spi/Kconfig +0 -1

index 169592a..f45d823 100644
@@ -76,7 +76,7 @@ config SPI_ATMEL
 
 config SPI_ATMEL_DMA
 	bool "Atmel SPI DMA support"
-	depends on SPI_ATMEL && (ARCH_AT91SAM9G45 || ARCH_AT91SAM9X5) &&

DMA_ENGINE && EXPERIMENTAL
+ depends on SPI_ATMEL && (SOC_AT91SAM9G45 || SOC_AT91SAM9X5) &&
DMA_ENGINE && EXPERIMENTAL
default y
help
Say Y here if you want the Atmel SPI driver to use the DMA engine.
Data transfers

Richard.

Signature · 1 line

for me, ck means con kolivas and not calvin klein... does it mean I'm a geek ?