threads / discuss / 17709

[PATCHv2] lis3lv02d: support both one- and two-byte sensors

Subject: [PATCHv2] lis3lv02d: support both one- and two-byte sensors

## tl;dr

4 messages between Feb 10, 2009 and Feb 10, 2009.

replies: 3people: 2as markdown or json

Giuseppe Bilotta· Feb 10, 2009, 23:40 UTC · lore

Sensors responding with 0x3B to WHO_AM_I only have one data register per direction, thus returning a signed byte from the position which is occupied by the MSB in sensors responding with 0x3A.

We support both kind of sensors by checking for the sensor type on init and defining appropriate data-access routines and sensor limits (for the joystick) depending on what we find.

Signed-off-by: Giuseppe Bilotta <giuseppe.bilotta@gmail.com>
---

This is the other version of the patch, and it changes access to use the base address from the 8-bit, so the 16-bit routine is changed to access the PREVIOUS byte for the LSB.

Choose whichever of the patches is deemed more correct 8-)

Note that both patches are based off Linus tree, so I don't know how they cope with Pavel's "don't touch too much on init" one. I'll try to work out a patch on top of that too.

 drivers/hwmon/lis3lv02d.c |   42 ++++++++++++++++++++++++++++++------------
 drivers/hwmon/lis3lv02d.h |   12 +++++++++++-
 2 files changed, 41 insertions(+), 13 deletions(-)
diff --git a/drivers/hwmon/lis3lv02d.c b/drivers/hwmon/lis3lv02d.c
index 219d2d0..b6b2d4a 100644
--- a/drivers/hwmon/lis3lv02d.c
+++ b/drivers/hwmon/lis3lv02d.c
@@ -52,9 +52,6 @@
  * joystick.
  */
 
-/* Maximum value our axis may get for the input device (signed 12 bits) */
-#define MDPS_MAX_VAL 2048
-
 struct acpi_lis3lv02d adev;
 EXPORT_SYMBOL_GPL(adev);
 
@@ -64,12 +61,19 @@ static s16 lis3lv02d_read_16(acpi_handle handle, int reg)
 {
 	u8 lo, hi;
 
-	adev.read(handle, reg, &lo);
-	adev.read(handle, reg + 1, &hi);
+	adev.read(handle, reg - 1, &lo);
+	adev.read(handle, reg, &hi);
 	/* In "12 bit right justified" mode, bit 6, bit 7, bit 8 = bit 5 */
 	return (s16)((hi << 8) | lo);
 }
 
+static s16 lis3lv02d_read_8(acpi_handle handle, int reg)
+{
+	u8 lo;
+	adev.read(handle, reg, &lo);
+	return *((s8*)(&lo));
+}
+
 /**
  * lis3lv02d_get_axis - For the given axis, give the value converted
  * @axis:      1,2,3 - can also be negative
@@ -98,9 +102,9 @@ static void lis3lv02d_get_xyz(acpi_handle handle, int *x, int *y, int *z)
 {
 	int position[3];
 
-	position[0] = lis3lv02d_read_16(handle, OUTX_L);
-	position[1] = lis3lv02d_read_16(handle, OUTY_L);
-	position[2] = lis3lv02d_read_16(handle, OUTZ_L);
+	position[0] = adev.lis3lv02d_read(handle, OUTX);
+	position[1] = adev.lis3lv02d_read(handle, OUTY);
+	position[2] = adev.lis3lv02d_read(handle, OUTZ);
 
 	*x = lis3lv02d_get_axis(adev.ac.x, position);
 	*y = lis3lv02d_get_axis(adev.ac.y, position);
@@ -121,7 +125,21 @@ void lis3lv02d_poweron(acpi_handle handle)
 
 	adev.is_on = 1;
 	adev.init(handle);
-	adev.write(handle, FF_WU_CFG, 0);
+	adev.read(handle, WHO_AM_I, &adev.whoami);
+	/* Tell apart LISxLV02Dy from LISx02Dy family by checking the LSB
+	 * (0x3A vs 0x3B). TODO More sophisticated checks on other registers
+	 * could be implemented, for example to see if we have 2 or 3 axes, and
+	 * configure the joystick accordingly.
+	 */
+	if (adev.whoami & 1) {
+		printk(KERN_INFO DRIVER_NAME ": 1-byte sensor found\n");
+		adev.lis3lv02d_read = lis3lv02d_read_8;
+		adev.mdps_max_val = 128;
+	} else {
+		printk(KERN_INFO DRIVER_NAME ": 2-byte sensor found\n");
+		adev.lis3lv02d_read = lis3lv02d_read_16;
+		adev.mdps_max_val = 2048;
+	}
 	/*
 	 * BDU: LSB and MSB values are not updated until both have been read.
 	 *      So the value read will always be correct.
@@ -231,9 +249,9 @@ int lis3lv02d_joystick_enable(void)
 	adev.idev->close      = lis3lv02d_joystick_close;
 
 	set_bit(EV_ABS, adev.idev->evbit);
-	input_set_abs_params(adev.idev, ABS_X, -MDPS_MAX_VAL, MDPS_MAX_VAL, 3, 3);
-	input_set_abs_params(adev.idev, ABS_Y, -MDPS_MAX_VAL, MDPS_MAX_VAL, 3, 3);
-	input_set_abs_params(adev.idev, ABS_Z, -MDPS_MAX_VAL, MDPS_MAX_VAL, 3, 3);
+	input_set_abs_params(adev.idev, ABS_X, -adev.mdps_max_val, adev.mdps_max_val, 3, 3);
+	input_set_abs_params(adev.idev, ABS_Y, -adev.mdps_max_val, adev.mdps_max_val, 3, 3);
+	input_set_abs_params(adev.idev, ABS_Z, -adev.mdps_max_val, adev.mdps_max_val, 3, 3);
 
 	err = input_register_device(adev.idev);
 	if (err) {
diff --git a/drivers/hwmon/lis3lv02d.h b/drivers/hwmon/lis3lv02d.h
index 223f1c0..9d7d4a8 100644
--- a/drivers/hwmon/lis3lv02d.h
+++ b/drivers/hwmon/lis3lv02d.h
@@ -22,11 +22,14 @@
 /*
  * The actual chip is STMicroelectronics LIS3LV02DL or LIS3LV02DQ that seems to
  * be connected via SPI. There exists also several similar chips (such as LIS302DL or
- * LIS3L02DQ) but not in the HP laptops and they have slightly different registers.
+ * LIS3L02DQ) and they have slightly different registers, but we can provide a
+ * common interface for all of them.
  * They can also be connected via I²C.
  */
 
+/* 2-byte registers */
 #define LIS3LV02DL_ID	0x3A /* Also the LIS3LV02DQ */
+/* 1-byte registers */
 #define LIS302DL_ID	0x3B /* Also the LIS202DL! */
 
 enum lis3lv02d_reg {
@@ -44,10 +47,13 @@ enum lis3lv02d_reg {
 	STATUS_REG	= 0x27,
 	OUTX_L		= 0x28,
 	OUTX_H		= 0x29,
+	OUTX		= 0x29,
 	OUTY_L		= 0x2A,
 	OUTY_H		= 0x2B,
+	OUTY		= 0x2B,
 	OUTZ_L		= 0x2C,
 	OUTZ_H		= 0x2D,
+	OUTZ		= 0x2D,
 	FF_WU_CFG	= 0x30,
 	FF_WU_SRC	= 0x31,
 	FF_WU_ACK	= 0x32,
@@ -159,6 +165,10 @@ struct acpi_lis3lv02d {
 	acpi_status (*write) (acpi_handle handle, int reg, u8 val);
 	acpi_status (*read) (acpi_handle handle, int reg, u8 *ret);
 
+	u8			whoami;    /* 3Ah: 2-byte registries, 3Bh: 1-byte registries */
+	s16 (*lis3lv02d_read) (acpi_handle handle, int reg);
+	int			mdps_max_val;
+
 	struct input_dev	*idev;     /* input device */
 	struct task_struct	*kthread;  /* kthread for input */
 	struct mutex            lock;
-- 
1.6.2.rc0.173.g5e148
Éric Piel· Feb 10, 2009, 23:46 UTC · re: Giuseppe Bilotta · lore

Re: [PATCHv2] lis3lv02d: support both one- and two-byte sensors

Giuseppe Bilotta schreef:
Show 20 quoted lines
> Sensors responding with 0x3B to WHO_AM_I only have one data register per
> direction, thus returning a signed byte from the position which is
> occupied by the MSB in sensors responding with 0x3A.
> 
> We support both kind of sensors by checking for the sensor type on init
> and defining appropriate data-access routines and sensor limits (for the
> joystick) depending on what we find.
> 
> Signed-off-by: Giuseppe Bilotta <giuseppe.bilotta@gmail.com>
> ---
> 
> This is the other version of the patch, and it changes access to use the
> base address from the 8-bit, so the 16-bit routine is changed to access
> the PREVIOUS byte for the LSB.
> 
> Choose whichever of the patches is deemed more correct 8-)
> 
> Note that both patches are based off Linus tree, so I don't know how
> they cope with Pavel's "don't touch too much on init" one. I'll try to
> work out a patch on top of that too.

Yes, great, and this one looks even better than v1 :-) However I'm not sure about the conversion between s8 and s16:

Show 6 quoted lines
> +static s16 lis3lv02d_read_8(acpi_handle handle, int reg)
> +{
> +	u8 lo;
> +	adev.read(handle, reg, &lo);
> +	return *((s8*)(&lo));
> +}

Does it really extend the sign to 16 bits? I would have written it this way: +static s16 lis3lv02d_read_8(acpi_handle handle, int reg) +{ + s8 lo; + adev.read(handle, reg, &lo); + return (s16)lo; +} Doesn't it work better?

Eric
Giuseppe Bilotta· Feb 10, 2009, 23:51 UTC · re: Éric Piel · lore

Re: [PATCHv2] lis3lv02d: support both one- and two-byte sensors

On Wed, Feb 11, 2009 at 12:46 AM, Éric Piel <Eric.Piel@tremplin-utc.net> wrote:
Show 24 quoted lines
> Giuseppe Bilotta schreef:
>>
>> Sensors responding with 0x3B to WHO_AM_I only have one data register per
>> direction, thus returning a signed byte from the position which is
>> occupied by the MSB in sensors responding with 0x3A.
>>
>> We support both kind of sensors by checking for the sensor type on init
>> and defining appropriate data-access routines and sensor limits (for the
>> joystick) depending on what we find.
>>
>> Signed-off-by: Giuseppe Bilotta <giuseppe.bilotta@gmail.com>
>> ---
>>
>> This is the other version of the patch, and it changes access to use the
>> base address from the 8-bit, so the 16-bit routine is changed to access
>> the PREVIOUS byte for the LSB.
>>
>> Choose whichever of the patches is deemed more correct 8-)
>>
>> Note that both patches are based off Linus tree, so I don't know how
>> they cope with Pavel's "don't touch too much on init" one. I'll try to
>> work out a patch on top of that too.
>
> Yes, great, and this one looks even better than v1 :-)
And don't forget about the v3 for -mm ;-)
Show 17 quoted lines
> However I'm not sure about the conversion between s8 and s16:
>
>> +static s16 lis3lv02d_read_8(acpi_handle handle, int reg)
>> +{
>> +       u8 lo;
>> +       adev.read(handle, reg, &lo);
>> +       return *((s8*)(&lo));
>> +}
>
> Does it really extend the sign to 16 bits? I would have written it this way:
> +static s16 lis3lv02d_read_8(acpi_handle handle, int reg)
> +{
> +       s8 lo;
> +       adev.read(handle, reg, &lo);
> +       return (s16)lo;
> +}
> Doesn't it work better?

Well, the previous one works, but I admit it's horribly convoluted. I'll double check with this one and resubmit (I'll only resubmit the one based on Pavel's "don't touch anything on init" probably though, hope it's ok)

-- 
Giuseppe "Oblomov" Bilotta
Éric Piel· Feb 10, 2009, 23:54 UTC · re: Giuseppe Bilotta · lore

Re: [PATCHv2] lis3lv02d: support both one- and two-byte sensors

Giuseppe Bilotta schreef:
> Well, the previous one works, but I admit it's horribly convoluted.
> I'll double check with this one and resubmit (I'll only resubmit the
> one based on Pavel's "don't touch anything on init" probably though,
> hope it's ok)

Sure, only modified v3 is ok. BTW, now that this is working. Could you double check the correct axes set up. In particular, is it x_inverted or y_inverted, and what about z axis...

Thanks, Eric

← back to recent threads