[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <20241031215749.10a8dbbe@jic23-huawei>
Date: Thu, 31 Oct 2024 21:57:49 +0000
From: Jonathan Cameron <jic23@...nel.org>
To: Jonathan Cameron <Jonathan.Cameron@...wei.com>
Cc: Matti Vaittinen <mazziesaccount@...il.com>, Matti Vaittinen
<matti.vaittinen@...rohmeurope.com>, Kalle Niemi <kaleposti@...il.com>,
"Lars-Peter Clausen" <lars@...afoo.de>, Andy Shevchenko
<andriy.shevchenko@...ux.intel.com>, <linux-iio@...r.kernel.org>,
<linux-kernel@...r.kernel.org>
Subject: Re: [PATCH] iio: kx022a: Fix raw read format
On Wed, 30 Oct 2024 18:06:58 +0000
Jonathan Cameron <Jonathan.Cameron@...wei.com> wrote:
> On Wed, 30 Oct 2024 15:16:11 +0200
> Matti Vaittinen <mazziesaccount@...il.com> wrote:
>
> > The KX022A provides the accelerometer data in two subsequent registers.
> > The registers are laid out so that the value obtained via bulk-read of
> > these registers can be interpreted as signed 16-bit little endian value.
> > The read value is converted to cpu_endianes and stored into 32bit integer.
> > The le16_to_cpu() casts value to unsigned 16-bit value, and when this is
> > assigned to 32-bit integer the resulting value will always be positive.
> >
> > This has not been a problem to users (at least not all users) of the sysfs
> > interface, who know the data format based on the scan info and who have
> > converted the read value back to 16-bit signed value.
> They shouldn't be doing that. Scaninfo is for buffered values only
> This should indeed be signed.
I added a note saying this isn't compliant with the ABI.
Applied to the fixes-togreg branch fo iio.git and marked for stable.
>
> >
> > This, however, will be a problem for those who use the in-kernel
> > interfaces, especially the iio_read_channel_processed_scale().
> >
> > The iio_read_channel_processed_scale() performs multiplications to the
> > returned (always positive) raw value, which will cause strange results
> > when the data from the sensor has been negative.
> >
> > Fix the read_raw format by casting the result of the le_to_cpu() to
> > signed 16-bit value before assigning it to the integer. This will make
> > the negative readings to be correctly reported as negative.
> >
> > This fix will be visible to users by changing values returned via sysfs
> > to appear in correct (negative) format.
> >
> > Reported-by: Kalle Niemi <kaleposti@...il.com>
> > Fixes: 7c1d1677b322 ("iio: accel: Support Kionix/ROHM KX022A accelerometer")
> > Signed-off-by: Matti Vaittinen <mazziesaccount@...il.com>
> > Tested-by: Kalle Niemi <kaleposti@...il.com>
> > ---
> > drivers/iio/accel/kionix-kx022a.c | 2 +-
> > 1 file changed, 1 insertion(+), 1 deletion(-)
> >
> > diff --git a/drivers/iio/accel/kionix-kx022a.c b/drivers/iio/accel/kionix-kx022a.c
> > index 53d59a04ae15..b6a828a6df93 100644
> > --- a/drivers/iio/accel/kionix-kx022a.c
> > +++ b/drivers/iio/accel/kionix-kx022a.c
> > @@ -594,7 +594,7 @@ static int kx022a_get_axis(struct kx022a_data *data,
> > if (ret)
> > return ret;
> >
> > - *val = le16_to_cpu(data->buffer[0]);
> > + *val = (s16)le16_to_cpu(data->buffer[0]);
> LGTM.
> >
> > return IIO_VAL_INT;
> > }
> >
> > base-commit: 81983758430957d9a5cb3333fe324fd70cf63e7e
>
Powered by blists - more mailing lists