Re: [PATCH v2 0/2] iio: magnetometer: add support for Infineon TLV493D 3D Magnetic Sensor

[Jonathan Cameron <jic23@kernel.org>]

On Sat, 2 Aug 2025 10:53:30 +0200
Andy Shevchenko <andy.shevchenko@gmail.com> wrote:

> On Sat, Aug 2, 2025 at 8:44 AM Dixit Parmar <dixitparmar19@gmail.com> wrote:
> >
> > The Infineon TLV493D is a Low-Power 3D Magnetic Sensor. The Sensor
> > applications includes joysticks, control elements (white goods,
> > multifunction knops), or electric meters (anti tampering) and any
> > other application that requires accurate angular measurements at
> > low power consumptions.
> >
> > The Sensor is configured over I2C, and as part of Sensor measurement
> > data it provides 3-Axis magnetic fields and temperature core measurement.
> >
> > The driver supports raw value read and buffered input via external trigger
> > to allow streaming values with the same sensing timestamp.
> >
> > While sensor has interrupt pin multiplexed with I2C SCL pin. But for bus  
> 
> the sensor
> an interrupt
> an I2C
> 
> > configurations interrupt(INT) is not recommended, unless timing constraints
> > between I2C data transfers and interrupt pulses are monitored and aligned.
> >
> > The Sensor's I2C register map and mode information is described in product
> > User Manual[Link].  
> 
> Replace Link here with 1...
> 
> > Datasheet: https://www.infineon.com/assets/row/public/documents/24/49/infineon-tlv493d-a1b6-datasheet-en.pdf
> > Link: https://www.mouser.com/pdfDocs/Infineon-TLV493D-A1B6_3DMagnetic-UserManual-v01_03-EN.pdf  
> 
> ...and add space followed by [1] here.
> 
> ...
> 
> > Changes in v2:
> > - Drop regmap implementation in favor of using direct i2c APIs to
> >   have uniform communication APIs across the driver.  
> 
> This I don't understand. I mean the motivation behind this. Usually
> direct I2C communication is used to do some initial checks and
> configuration and rarely for the actuall run-time driver
> functionality. Otherwise it means that the regmap may be used with a
> customised read and write methods.
> 

This was my suggestion.  The device has very odd characteristics that
means writes really are not register based.  You have to write them all
every time with now addressing.

So to me regmap just wasn't a good choice here. You could do something
some nasty stuff to hammer it into a custom regmap, but to me it's just
not a good fit.

Jonathan

>