diff mbox series

[v1] iio: gyro: mpu3050: Fix reported temperature value

Message ID 20210421234820.32211-1-digetx@gmail.com (mailing list archive)
State New, archived
Headers show
Series [v1] iio: gyro: mpu3050: Fix reported temperature value | expand

Commit Message

Dmitry Osipenko April 21, 2021, 11:48 p.m. UTC
The raw temperature value is a signed 16bit integer. The sign casting
is missed in the code, which results in a wrong temperature reported by
userspace tools, fix it.

Cc: stable@vger.kernel.org
Link: https://www.cdiweb.com/datasheets/invensense/mpu-3000a.pdf
Tested-by: Maxim Schwalm <maxim.schwalm@gmail.com> # Asus TF700T
Tested-by: Svyatoslav Ryhel <clamor95@gmail.com> # Asus TF201
Reported-by: Svyatoslav Ryhel <clamor95@gmail.com>
Signed-off-by: Dmitry Osipenko <digetx@gmail.com>
---
 drivers/iio/gyro/mpu3050-core.c | 13 +++++++++++--
 1 file changed, 11 insertions(+), 2 deletions(-)

Comments

Linus Walleij April 22, 2021, 12:38 a.m. UTC | #1
On Thu, Apr 22, 2021 at 1:49 AM Dmitry Osipenko <digetx@gmail.com> wrote:

> The raw temperature value is a signed 16bit integer. The sign casting
> is missed in the code, which results in a wrong temperature reported by
> userspace tools, fix it.
>
> Cc: stable@vger.kernel.org
> Link: https://www.cdiweb.com/datasheets/invensense/mpu-3000a.pdf
> Tested-by: Maxim Schwalm <maxim.schwalm@gmail.com> # Asus TF700T
> Tested-by: Svyatoslav Ryhel <clamor95@gmail.com> # Asus TF201
> Reported-by: Svyatoslav Ryhel <clamor95@gmail.com>
> Signed-off-by: Dmitry Osipenko <digetx@gmail.com>

+/- Andy's comments:
Reviewed-by: Linus Walleij <linus.walleij@linaro.org>

I never thought this driver would have so many users (3 people signed
testing it!) but I realize it is more widely deployed than I thought.

I have totally ignored the MPU3050's ability to act as a "sensor hub"
and talk to accelerometers and magnetometers directly. I always
thought it would be better to just route the I2C right through it and
put Linux in direct control, but I realize this was not Invensese's
intention. I don't know if it can be actually utilized in some generic
way, all kernels using that have separate hacky drivers for all the
sub-sensors duplicating the kernel drivers we already have ...

Yours,
Linus Walleij
Dmitry Osipenko April 22, 2021, 1:10 a.m. UTC | #2
22.04.2021 03:38, Linus Walleij пишет:
> On Thu, Apr 22, 2021 at 1:49 AM Dmitry Osipenko <digetx@gmail.com> wrote:
> 
>> The raw temperature value is a signed 16bit integer. The sign casting
>> is missed in the code, which results in a wrong temperature reported by
>> userspace tools, fix it.
>>
>> Cc: stable@vger.kernel.org
>> Link: https://www.cdiweb.com/datasheets/invensense/mpu-3000a.pdf
>> Tested-by: Maxim Schwalm <maxim.schwalm@gmail.com> # Asus TF700T
>> Tested-by: Svyatoslav Ryhel <clamor95@gmail.com> # Asus TF201
>> Reported-by: Svyatoslav Ryhel <clamor95@gmail.com>
>> Signed-off-by: Dmitry Osipenko <digetx@gmail.com>
> 
> +/- Andy's comments:
> Reviewed-by: Linus Walleij <linus.walleij@linaro.org>
> 
> I never thought this driver would have so many users (3 people signed
> testing it!) but I realize it is more widely deployed than I thought.
> 
> I have totally ignored the MPU3050's ability to act as a "sensor hub"
> and talk to accelerometers and magnetometers directly. I always
> thought it would be better to just route the I2C right through it and
> put Linux in direct control, but I realize this was not Invensese's
> intention. I don't know if it can be actually utilized in some generic
> way, all kernels using that have separate hacky drivers for all the
> sub-sensors duplicating the kernel drivers we already have ...

I don't think that MPU3050 could talk to the sensors behind it directly.
It's has "I2C gate" which allows to route the I2C communication to the
chained sensors, which is done in order to reduce noise on the I2C bus
such that only one sensor is active at a time. Those chained sensors are
working good using upstream kernel drivers, the accelerometer is
particularly useful for display autorotation. Modern DEs like Gnome and
KDE are using iio-sensor-proxy library that knows how to work with the
mainline sensor drivers.

Thank you and Andy for the review, I'll prepare v2.
Dmitry Osipenko April 22, 2021, 1:20 a.m. UTC | #3
22.04.2021 03:34, Andy Shevchenko пишет:
>     -                       /* The temperature scaling is (x+23000)/280
>     Celsius */
>     +                       /*
>     +                        * The temperature scaling is (x+23000)/280
>     Celsius,
>     +                        * where 23000 includes room temperature
>     offset of
>     +                        * +35C, 280 is the precision scale and x is
>     the signed
>     +                        * 16bit integer which corresponds to the
>     temperature
> 
> 
> “...16-bit signed integer...”
>  
> 
>     +                        * range of -40C..85C.
> 
> 
> 
> Datasheet says -30°C..+85°C.

Datasheet says this for the "Best fit straight line", while the min
value is -40C.
diff mbox series

Patch

diff --git a/drivers/iio/gyro/mpu3050-core.c b/drivers/iio/gyro/mpu3050-core.c
index ac90be03332a..9885428ca12e 100644
--- a/drivers/iio/gyro/mpu3050-core.c
+++ b/drivers/iio/gyro/mpu3050-core.c
@@ -272,7 +272,16 @@  static int mpu3050_read_raw(struct iio_dev *indio_dev,
 	case IIO_CHAN_INFO_OFFSET:
 		switch (chan->type) {
 		case IIO_TEMP:
-			/* The temperature scaling is (x+23000)/280 Celsius */
+			/*
+			 * The temperature scaling is (x+23000)/280 Celsius,
+			 * where 23000 includes room temperature offset of
+			 * +35C, 280 is the precision scale and x is the signed
+			 * 16bit integer which corresponds to the temperature
+			 * range of -40C..85C.
+			 *
+			 * Temperature value itself represents temperature of
+			 * the sensor die.
+			 */
 			*val = 23000;
 			return IIO_VAL_INT;
 		default:
@@ -329,7 +338,7 @@  static int mpu3050_read_raw(struct iio_dev *indio_dev,
 				goto out_read_raw_unlock;
 			}
 
-			*val = be16_to_cpu(raw_val);
+			*val = (s16)be16_to_cpu(raw_val);
 			ret = IIO_VAL_INT;
 
 			goto out_read_raw_unlock;