[PATCH v3] iio: mma8452: support either of the available interrupt pins

[PATCH v3] iio: mma8452: support either of the available interrupt pins

[Martin Kepplinger]

This change is important in order for everyone to be easily able to use the
driver for one of the supported accelerometer chips!

Until now, the driver blindly assumed that the INT1 interrupt line is wired
on a user's board. But these devices have 2 interrupt lines and can route
their different interrupt sources to one of them. Since only their motion
detection interrupt source is implemented as IIO events, users just use
either one of the pins.

Of course, this also falls back to assuming INT1, so for existing users
nothing will break. The new functionality is described in the bindings doc.

Signed-off-by: Martin Kepplinger <martin.kepplinger@theobroma-systems.com>
---
changelog:
v3: correctly assign irq if both pins are described in DT. thanks again
v2: don't warn but normally handle if both pins are described in dts
    thanks Mark Rutland
v1: initial post

 .../devicetree/bindings/iio/accel/mma8452.txt      |  6 +++++
 drivers/iio/accel/mma8452.c                        | 26 +++++++++++++++++-----
 2 files changed, 26 insertions(+), 6 deletions(-)

diff --git a/Documentation/devicetree/bindings/iio/accel/mma8452.txt b/Documentation/devicetree/bindings/iio/accel/mma8452.txt
index e3c3746..3c10e85 100644
--- a/Documentation/devicetree/bindings/iio/accel/mma8452.txt
+++ b/Documentation/devicetree/bindings/iio/accel/mma8452.txt
@@ -7,13 +7,18 @@ Required properties:
     * "fsl,mma8453"
     * "fsl,mma8652"
     * "fsl,mma8653"
+
   - reg: the I2C address of the chip
 
 Optional properties:
 
   - interrupt-parent: should be the phandle for the interrupt controller
+
   - interrupts: interrupt mapping for GPIO IRQ
 
+  - interrupt-names: should contain "INT1" and/or "INT2", the accelerometer's
+		     interrupt line in use.
+
 Example:
 
 	mma8453fc@1d {
@@ -21,4 +26,5 @@ Example:
 		reg = <0x1d>;
 		interrupt-parent = <&gpio1>;
 		interrupts = <5 0>;
+		interrupt-names = "INT2";
 	};
diff --git a/drivers/iio/accel/mma8452.c b/drivers/iio/accel/mma8452.c
index 1eccc2d..6b9b360 100644
--- a/drivers/iio/accel/mma8452.c
+++ b/drivers/iio/accel/mma8452.c
@@ -29,6 +29,7 @@
 #include <linux/iio/events.h>
 #include <linux/delay.h>
 #include <linux/of_device.h>
+#include <linux/of_irq.h>
 
 #define MMA8452_STATUS				0x00
 #define  MMA8452_STATUS_DRDY			(BIT(2) | BIT(1) | BIT(0))
@@ -1130,13 +1131,26 @@ static int mma8452_probe(struct i2c_client *client,
 					   MMA8452_INT_FF_MT;
 		int enabled_interrupts = MMA8452_INT_TRANS |
 					 MMA8452_INT_FF_MT;
+		int irq1, irq2;
 
-		/* Assume wired to INT1 pin */
-		ret = i2c_smbus_write_byte_data(client,
-						MMA8452_CTRL_REG5,
-						supported_interrupts);
-		if (ret < 0)
-			return ret;
+		irq1 = of_irq_get_byname(client->dev.of_node, "INT1");
+		irq2 = of_irq_get_byname(client->dev.of_node, "INT2");
+
+		/* if INT2 is found, use it. Otherwise INT1 */
+		if (!(irq2 > 0 && irq1 < 0)) {
+			ret = i2c_smbus_write_byte_data(client,
+							MMA8452_CTRL_REG5,
+							supported_interrupts);
+			if (ret < 0)
+				return ret;
+
+			if (irq1 > 0)
+				client->irq = irq1;
+			dev_info(&client->dev, "using interrupt line INT1\n");
+		} else {
+			client->irq = irq2;
+			dev_info(&client->dev, "using interrupt line INT2\n");
+		}
 
 		ret = i2c_smbus_write_byte_data(client,
 						MMA8452_CTRL_REG4,
-- 
2.1.4

Re: [PATCH v3] iio: mma8452: support either of the available interrupt pins

[Lars-Peter Clausen]

On 10/14/2015 03:15 PM, Martin Kepplinger wrote:
[...]
> +			if (irq1 > 0)
> +				client->irq = irq1;

You must not overwrite client->irq, that field is manged by the I2C core and
is supposed to be read only for device drivers.

> +			dev_info(&client->dev, "using interrupt line INT1\n");

This should probably be a dev_dbg()

> +		} else {
> +			client->irq = irq2;
> +			dev_info(&client->dev, "using interrupt line INT2\n");

Same here

> +		}

Re: [PATCH v3] iio: mma8452: support either of the available interrupt pins

[Lars-Peter Clausen]

On 10/14/2015 03:15 PM, Martin Kepplinger wrote:
[...]
> +			if (irq1 > 0)
> +				client->irq = irq1;

You must not overwrite client->irq, that field is manged by the I2C core and
is supposed to be read only for device drivers.

> +			dev_info(&client->dev, "using interrupt line INT1\n");

This should probably be a dev_dbg()

> +		} else {
> +			client->irq = irq2;
> +			dev_info(&client->dev, "using interrupt line INT2\n");

Same here

> +		}

Re: [PATCH v3] iio: mma8452: support either of the available interrupt pins

[Martin Kepplinger]

Am 2015-10-14 um 17:12 schrieb Lars-Peter Clausen:
> On 10/14/2015 03:15 PM, Martin Kepplinger wrote:
> [...]
>> +			if (irq1 > 0)
>> +				client->irq = irq1;
> 
> You must not overwrite client->irq, that field is manged by the I2C core and
> is supposed to be read only for device drivers.
>

I thought about it again and before I implement it, let me show you:

since interrupt-names would be "invented" anyways ("INT1", "INT2"),
here's an idea for the bindings doc that would be a more long-term
solution to implement:

  - interrupts: interrupt mapping for GPIO IRQ

        These devices have two interrupt pins called INT1 and INT2 they
        can route their different interrupt sources to:

        IRQ Name        Interrupt Source        Wired Pin
        --------        ----------------        ---------
        DATA_READY_1    data ready              INT1
        DATA_READY_2                            INT2
        MOTION_1        motion events           INT1
        MOTION_2                                INT2
        INT1            all                     INT1
        INT2                                    INT2

  - interrupt-names: should contain IRQ Names in the order in which they
                     were supplied in the interrupts property.

                     Depending on how your chip is wired and what
                     interrupt sources should be handled by the driver,
                     choose one IRQ Name per Interrupt source, or
                     INT1/INT2 for all sources to one pin here.

Example 1:

        mma8453fc@1d {
                compatible = "fsl,mma8453";
                reg = <0x1d>;
                interrupt-parent = <&gpio1>;
                interrupts = <5 0>;
                interrupt-names = "INT2"; // interrupt pin wired to INT2
        };

Example 2:

        mma8453fc@1d {
                compatible = "fsl,mma8453";
                reg = <0x1d>;
                interrupt-parent = <&gpio1>;
                interrupts = <4 0>, <5 0>;
                /* data ready comes from INT2, motion events from INT1 */
                interrupt-names = "DATA_READY_2", "MOTION_1";
        };

What do you think?

thanks
                           martin

Re: [PATCH v3] iio: mma8452: support either of the available interrupt pins

[Martin Kepplinger]

Am 2015-10-14 um 17:12 schrieb Lars-Peter Clausen:
> On 10/14/2015 03:15 PM, Martin Kepplinger wrote:
> [...]
>> +			if (irq1 > 0)
>> +				client->irq = irq1;
> 
> You must not overwrite client->irq, that field is manged by the I2C core and
> is supposed to be read only for device drivers.
>

I thought about it again and before I implement it, let me show you:

since interrupt-names would be "invented" anyways ("INT1", "INT2"),
here's an idea for the bindings doc that would be a more long-term
solution to implement:

  - interrupts: interrupt mapping for GPIO IRQ

        These devices have two interrupt pins called INT1 and INT2 they
        can route their different interrupt sources to:

        IRQ Name        Interrupt Source        Wired Pin
        --------        ----------------        ---------
        DATA_READY_1    data ready              INT1
        DATA_READY_2                            INT2
        MOTION_1        motion events           INT1
        MOTION_2                                INT2
        INT1            all                     INT1
        INT2                                    INT2

  - interrupt-names: should contain IRQ Names in the order in which they
                     were supplied in the interrupts property.

                     Depending on how your chip is wired and what
                     interrupt sources should be handled by the driver,
                     choose one IRQ Name per Interrupt source, or
                     INT1/INT2 for all sources to one pin here.

Example 1:

        mma8453fc@1d {
                compatible = "fsl,mma8453";
                reg = <0x1d>;
                interrupt-parent = <&gpio1>;
                interrupts = <5 0>;
                interrupt-names = "INT2"; // interrupt pin wired to INT2
        };

Example 2:

        mma8453fc@1d {
                compatible = "fsl,mma8453";
                reg = <0x1d>;
                interrupt-parent = <&gpio1>;
                interrupts = <4 0>, <5 0>;
                /* data ready comes from INT2, motion events from INT1 */
                interrupt-names = "DATA_READY_2", "MOTION_1";
        };

What do you think?

thanks
                           martin

Re: [PATCH v3] iio: mma8452: support either of the available interrupt pins

[Mark Rutland]

On Thu, Oct 15, 2015 at 11:32:59AM +0200, Martin Kepplinger wrote:
> Am 2015-10-14 um 17:12 schrieb Lars-Peter Clausen:
> > On 10/14/2015 03:15 PM, Martin Kepplinger wrote:
> > [...]
> >> +			if (irq1 > 0)
> >> +				client->irq = irq1;
> > 
> > You must not overwrite client->irq, that field is manged by the I2C core and
> > is supposed to be read only for device drivers.
> >
> 
> I thought about it again and before I implement it, let me show you:
> 
> since interrupt-names would be "invented" anyways ("INT1", "INT2"),
> here's an idea for the bindings doc that would be a more long-term
> solution to implement:
> 
>   - interrupts: interrupt mapping for GPIO IRQ
> 
>         These devices have two interrupt pins called INT1 and INT2 they
>         can route their different interrupt sources to:

This being the case, the binding should only talk about INT1 and INT2.
The names might be "invented", but they describe the physical pins on
the device, and thus describe a physical property of the device.

>         IRQ Name        Interrupt Source        Wired Pin
>         --------        ----------------        ---------
>         DATA_READY_1    data ready              INT1
>         DATA_READY_2                            INT2
>         MOTION_1        motion events           INT1
>         MOTION_2                                INT2
>         INT1            all                     INT1
>         INT2                                    INT2
> 
>   - interrupt-names: should contain IRQ Names in the order in which they
>                      were supplied in the interrupts property.
> 
>                      Depending on how your chip is wired and what
>                      interrupt sources should be handled by the driver,
>                      choose one IRQ Name per Interrupt source, or
>                      INT1/INT2 for all sources to one pin here.

The configuration of logical interrupts to those physical pins is a
choice that can be made at runtime, and should not live in the DT. It is
not a property of the hardware.

Thanks,
Mark.

Re: [PATCH v3] iio: mma8452: support either of the available interrupt pins

[Mark Rutland]

On Thu, Oct 15, 2015 at 11:32:59AM +0200, Martin Kepplinger wrote:
> Am 2015-10-14 um 17:12 schrieb Lars-Peter Clausen:
> > On 10/14/2015 03:15 PM, Martin Kepplinger wrote:
> > [...]
> >> +			if (irq1 > 0)
> >> +				client->irq = irq1;
> > 
> > You must not overwrite client->irq, that field is manged by the I2C core and
> > is supposed to be read only for device drivers.
> >
> 
> I thought about it again and before I implement it, let me show you:
> 
> since interrupt-names would be "invented" anyways ("INT1", "INT2"),
> here's an idea for the bindings doc that would be a more long-term
> solution to implement:
> 
>   - interrupts: interrupt mapping for GPIO IRQ
> 
>         These devices have two interrupt pins called INT1 and INT2 they
>         can route their different interrupt sources to:

This being the case, the binding should only talk about INT1 and INT2.
The names might be "invented", but they describe the physical pins on
the device, and thus describe a physical property of the device.

>         IRQ Name        Interrupt Source        Wired Pin
>         --------        ----------------        ---------
>         DATA_READY_1    data ready              INT1
>         DATA_READY_2                            INT2
>         MOTION_1        motion events           INT1
>         MOTION_2                                INT2
>         INT1            all                     INT1
>         INT2                                    INT2
> 
>   - interrupt-names: should contain IRQ Names in the order in which they
>                      were supplied in the interrupts property.
> 
>                      Depending on how your chip is wired and what
>                      interrupt sources should be handled by the driver,
>                      choose one IRQ Name per Interrupt source, or
>                      INT1/INT2 for all sources to one pin here.

The configuration of logical interrupts to those physical pins is a
choice that can be made at runtime, and should not live in the DT. It is
not a property of the hardware.

Thanks,
Mark.
--
To unsubscribe from this list: send the line "unsubscribe devicetree" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: [PATCH v3] iio: mma8452: support either of the available interrupt pins

[Mark Rutland]

On Wed, Oct 14, 2015 at 05:12:32PM +0200, Lars-Peter Clausen wrote:
> On 10/14/2015 03:15 PM, Martin Kepplinger wrote:
> [...]
> > +			if (irq1 > 0)
> > +				client->irq = irq1;
> 
> You must not overwrite client->irq, that field is manged by the I2C core and
> is supposed to be read only for device drivers.

Ok.

I think for the moment the simlpest thing to do would be to check which
of INT1 and INT2 match client->irq, and use that.

Thanks,
Mark.

Re: [PATCH v3] iio: mma8452: support either of the available interrupt pins

[Mark Rutland]

On Wed, Oct 14, 2015 at 05:12:32PM +0200, Lars-Peter Clausen wrote:
> On 10/14/2015 03:15 PM, Martin Kepplinger wrote:
> [...]
> > +			if (irq1 > 0)
> > +				client->irq = irq1;
> 
> You must not overwrite client->irq, that field is manged by the I2C core and
> is supposed to be read only for device drivers.

Ok.

I think for the moment the simlpest thing to do would be to check which
of INT1 and INT2 match client->irq, and use that.

Thanks,
Mark.