Re: [PATCH] i2c-designware: Interrupt sharing and SDA hold time

[Mika Westerberg <mika.westerberg@linux.intel.com>]

On Fri, Feb 15, 2013 at 02:05:10PM +0100, Christian Ruppert wrote:
> On Fri, Feb 15, 2013 at 01:50:37PM +0200, Mika Westerberg wrote:
> > On Fri, Feb 15, 2013 at 11:11:50AM +0100, Christian Ruppert wrote:
> > > This patch implements interrupt sharing and SDA hold time configuration in
> > > the designware i2c driver. Both functions are enabled through platform data
> > > or device tree. Tested with Linux-3.8rc4 (Synopsys ARC port, see
> > > https://github.com/foss-for-synopsys-dwc-arc-processors/linux/commits/arc-3.8-baseline).
> > > 
> > > Signed-off-by: Christian Ruppert <christian.ruppert@abilis.com>
> > > Signed-off-by: Pierrick Hascoet <pierrick.hascoet@abilis.com>
> > > ---
> > >  .../devicetree/bindings/i2c/i2c-designware.txt     |   18 ++++++++++
> > >  drivers/i2c/busses/i2c-designware-core.c           |    6 +++
> > >  drivers/i2c/busses/i2c-designware-core.h           |    2 +
> > >  drivers/i2c/busses/i2c-designware-platdrv.c        |   34 +++++++++++++++++++-
> > >  include/linux/i2c/i2c-designware.h                 |   13 +++++++
> > >  5 files changed, 72 insertions(+), 1 deletions(-)
> > >  create mode 100644 include/linux/i2c/i2c-designware.h
> > > 
> > > diff --git a/Documentation/devicetree/bindings/i2c/i2c-designware.txt b/Documentation/devicetree/bindings/i2c/i2c-designware.txt
> > > index e42a2ee..474386b 100644
> > > --- a/Documentation/devicetree/bindings/i2c/i2c-designware.txt
> > > +++ b/Documentation/devicetree/bindings/i2c/i2c-designware.txt
> > > @@ -10,6 +10,12 @@ Recommended properties :
> > >  
> > >   - clock-frequency : desired I2C bus clock frequency in Hz.
> > >  
> > > +Optional properties :
> > > + - interrupts-shared : If this property is present, the interrupt line is
> > > +   considered shared with other devices.
> > > + - sda-hold-time : If this property is present, the register SDA_HOLD will
> > > +   be initialised with its value.
> > > +
> > >  Example :
> > >  
> > >  	i2c@f0000 {
> > > @@ -20,3 +26,15 @@ Example :
> > >  		interrupts = <11>;
> > >  		clock-frequency = <400000>;
> > >  	};
> > > +
> > > +	i2c@1120000 {
> > > +		#address-cells = <1>;
> > > +		#size-cells = <0>;
> > > +		compatible = "snps,designware-i2c";
> > > +		reg = <0x1120000 0x1000>;
> > > +		interrupt-parent = <&ictl>;
> > > +		interrupts = <12 1>;
> > > +		interrupts-shared;
> > > +		clock-frequency = <400000>;
> > > +		sda-hold-time = <0x64>;
> > > +	};
> > > diff --git a/drivers/i2c/busses/i2c-designware-core.c b/drivers/i2c/busses/i2c-designware-core.c
> > > index cbba7db..36ba0b5 100644
> > > --- a/drivers/i2c/busses/i2c-designware-core.c
> > > +++ b/drivers/i2c/busses/i2c-designware-core.c
> > > @@ -34,6 +34,7 @@
> > >  #include <linux/io.h>
> > >  #include <linux/pm_runtime.h>
> > >  #include <linux/delay.h>
> > > +#include <linux/i2c/i2c-designware.h>
> > >  #include "i2c-designware-core.h"
> > >  
> > >  /*
> > > @@ -66,6 +67,7 @@
> > >  #define DW_IC_STATUS		0x70
> > >  #define DW_IC_TXFLR		0x74
> > >  #define DW_IC_RXFLR		0x78
> > > +#define DW_IC_SDA_HOLD		0x7c
> > >  #define DW_IC_TX_ABRT_SOURCE	0x80
> > >  #define DW_IC_COMP_PARAM_1	0xf4
> > >  #define DW_IC_COMP_TYPE		0xfc
> > > @@ -309,6 +311,10 @@ int i2c_dw_init(struct dw_i2c_dev *dev)
> > >  	dw_writel(dev, lcnt, DW_IC_FS_SCL_LCNT);
> > >  	dev_dbg(dev->dev, "Fast-mode HCNT:LCNT = %d:%d\n", hcnt, lcnt);
> > >  
> > > +	/* Configure SDA Hold Time if required */
> > > +	if (dev->pdat_flags & I2C_DW_PDAT_FLAGS_SET_HOLDTIME)
> > > +		dw_writel(dev, dev->sda_hold_time, DW_IC_SDA_HOLD);
> > 
> > AFAIK, DW_IC_SDA_HOLD minimum value is 1 so you could just do:
> > 
> > 	if (dev->sda_hold_time)
> > 		dw_writel(dev, dev->sda_hold_time, DW_IC_SDA_HOLD);
> 
> This is correct, I'll modify it.
> 
> > Also should this be configurable for standard and fast modes both?
> > Something like sm_sda_hold_time and fm_sda_hold_time?
> 
> AFAIK the requirements for data hold time do not differ between standard
> and fast mode. The hardware seems to have been designed in the same
> understanding since it provides two separate register sets for clock
> timings in standard and fast modes but only one register for data hold
> time. Am I missing something here?

No, you are probably right. We can always add support for the other mode in
the driver if there is need for that.

> > > +
> > >  	/* Configure Tx/Rx FIFO threshold levels */
> > >  	dw_writel(dev, dev->tx_fifo_depth - 1, DW_IC_TX_TL);
> > >  	dw_writel(dev, 0, DW_IC_RX_TL);
> > > diff --git a/drivers/i2c/busses/i2c-designware-core.h b/drivers/i2c/busses/i2c-designware-core.h
> > > index 9c1840e..ef20477 100644
> > > --- a/drivers/i2c/busses/i2c-designware-core.h
> > > +++ b/drivers/i2c/busses/i2c-designware-core.h
> > > @@ -88,6 +88,8 @@ struct dw_i2c_dev {
> > >  	u32			master_cfg;
> > >  	unsigned int		tx_fifo_depth;
> > >  	unsigned int		rx_fifo_depth;
> > > +	u32			sda_hold_time;
> > > +	unsigned int		pdat_flags;
> > >  };
> > >  
> > >  #define ACCESS_SWAP		0x00000001
> > > diff --git a/drivers/i2c/busses/i2c-designware-platdrv.c b/drivers/i2c/busses/i2c-designware-platdrv.c
> > > index 343357a..2ad2da3 100644
> > > --- a/drivers/i2c/busses/i2c-designware-platdrv.c
> > > +++ b/drivers/i2c/busses/i2c-designware-platdrv.c
> > > @@ -34,11 +34,13 @@
> > >  #include <linux/sched.h>
> > >  #include <linux/err.h>
> > >  #include <linux/interrupt.h>
> > > +#include <linux/of.h>
> > >  #include <linux/of_i2c.h>
> > >  #include <linux/platform_device.h>
> > >  #include <linux/pm.h>
> > >  #include <linux/io.h>
> > >  #include <linux/slab.h>
> > > +#include <linux/i2c/i2c-designware.h>
> > >  #include "i2c-designware-core.h"
> > >  
> > >  static struct i2c_algorithm i2c_dw_algo = {
> > > @@ -56,6 +58,28 @@ static int dw_i2c_probe(struct platform_device *pdev)
> > >  	struct i2c_adapter *adap;
> > >  	struct resource *mem, *ioarea;
> > >  	int irq, r;
> > > +	unsigned long irqf;
> > > +	struct i2c_designware_platform_data *pdat = pdev->dev.platform_data;
> > > +
> > > +	if ((!pdat) && pdev->dev.of_node) {
> > > +		struct device_node *node = pdev->dev.of_node;
> > > +
> > > +		pdat = kmalloc(sizeof(struct i2c_designware_platform_data),
> > > +				GFP_KERNEL);
> > > +		if (!pdat)
> > > +			return -ENOMEM;
> > > +
> > > +		pdat->flags = 0;
> > > +
> > > +		if (of_get_property(node, "interrupts-shared", NULL))
> > > +			pdat->flags |= I2C_DW_PDAT_FLAGS_SHARE_IRQ;
> > > +
> > > +		if (!of_property_read_u32(node, "sda-hold-time",
> > > +					&pdat->sda_hold_time))
> > > +			pdat->flags |= I2C_DW_PDAT_FLAGS_SET_HOLDTIME;
> > > +
> > > +		pdev->dev.platform_data = pdat;
> > 
> > I do not understand why you need a separate platform data at all? Can't you
> > do it in a same way than the ACPI code?
> 
> I suppose you are referring to dw_i2c_acpi_configure() in i2c/next which
> directly sets the values dev->... This is perfectly feasible and would
> greatly simplify the patch. The reason I introduced the platform data
> indirection layer is compatibility with legacy instantiations and
> platforms that do not yet support device tree. We do not internally need
> that backwards compatibility but does anybody else? Otherwise I'll
> remove it.

Well, there wasn't any platform data before so I'm not sure if it breaks
anything. And if you only instantiate the new fields once you know that the
device has been enumerated from DT, you should be fine.

> 
> > And this probably doesn't apply to the i2c/next branch because we added the
> > ACPI enumeration there.
> 
> No problem, I'll check the next patch iteration wrt. i2c/next which I
> should have probably done in the first place.

Cool.

> 
> > > +	}
> > >  
> > >  	/* NOTE: driver uses the static register mapping */
> > >  	mem = platform_get_resource(pdev, IORESOURCE_MEM, 0);
> > > @@ -120,12 +144,20 @@ static int dw_i2c_probe(struct platform_device *pdev)
> > >  		dev->tx_fifo_depth = ((param1 >> 16) & 0xff) + 1;
> > >  		dev->rx_fifo_depth = ((param1 >> 8)  & 0xff) + 1;
> > >  	}
> > > +	dev->sda_hold_time = pdat->sda_hold_time;
> > > +	dev->pdat_flags = pdat->flags;
> > > +
> > >  	r = i2c_dw_init(dev);
> > >  	if (r)
> > >  		goto err_iounmap;
> > >  
> > >  	i2c_dw_disable_int(dev);
> > > -	r = request_irq(dev->irq, i2c_dw_isr, IRQF_DISABLED, pdev->name, dev);
> > > +
> > > +	irqf = IRQF_DISABLED;
> > > +	if (dev->pdat_flags & I2C_DW_PDAT_FLAGS_SHARE_IRQ)
> > > +		irqf |= IRQF_SHARED;
> > 
> > Please check Wolfram's i2c/next. This is set already there.
> 
> Perfect. My fault for not looking there first. Thanks for highlighting
> this.

No problem :)

And thanks for adding SDA_HOLD support - we also have need for that on
Lynxpoint platform.