Message-ID: <20171210174826.265a3329@archlinux>
Date:   Sun, 10 Dec 2017 17:48:26 +0000
From:   Jonathan Cameron <jic23@...nel.org>
To:     Andreas Färber <afaerber@...e.de>
Cc:     Peter Meerwald-Stadler <pmeerw@...erw.net>,
        linux-iio@...r.kernel.org, linux-kernel@...r.kernel.org,
        Marius Tarcatu <marius.tarcatu@...ineon.com>,
        Hartmut Knaack <knaack.h@....de>,
        Lars-Peter Clausen <lars@...afoo.de>
Subject: Re: [PATCH 2/2] iio: magnetometer: Add Infineon TLV493D-A1B6

On Sun, 10 Dec 2017 18:14:44 +0100
Andreas Färber <afaerber@...e.de> wrote:

> Hi,
Some quick notes inline.

Jonathan

> 
> Am 10.12.2017 um 10:08 schrieb Peter Meerwald-Stadler:
> >> The Infineon TLV493D is a 3D magnetic sensor, A1B6 is I2C based.  
> > 
> > comments below
> >    
> >> It is found among others on the Infineon 3D Magnetic Sensor 2Go Kit,
> >> which allows to detach the sensor as a breakout board.
> >>
> >> Cc: Marius Tarcatu <marius.tarcatu@...ineon.com>
> >> Signed-off-by: Andreas Färber <afaerber@...e.de>
> >> ---
> >>  drivers/iio/magnetometer/Kconfig   |   9 +
> >>  drivers/iio/magnetometer/Makefile  |   2 +
> >>  drivers/iio/magnetometer/tlv493d.c | 328 +++++++++++++++++++++++++++++++++++++
> >>  3 files changed, 339 insertions(+)
> >>  create mode 100644 drivers/iio/magnetometer/tlv493d.c
> >>
> >> diff --git a/drivers/iio/magnetometer/Kconfig b/drivers/iio/magnetometer/Kconfig
> >> index ed9d776d01af..5945d88a1595 100644
> >> --- a/drivers/iio/magnetometer/Kconfig
> >> +++ b/drivers/iio/magnetometer/Kconfig
> >> @@ -175,4 +175,13 @@ config SENSORS_HMC5843_SPI
> >>  	  - hmc5843_core (core functions)
> >>  	  - hmc5843_spi (support for HMC5983)
> >>  
> >> +config TLV493D
> >> +	tristate "Infineon TLV493D 3D Magnetic Sensor"
> >> +	depends on I2C
> >> +	help
> >> +	  Say Y here to build support for Infineon TLV493D-A1B6 I2C-based
> >> +	  3-axis magnetometer.
> >> +
> >> +	  This driver can also be built as a module and will be called tlv493d.
> >> +
> >>  endmenu
> >> diff --git a/drivers/iio/magnetometer/Makefile b/drivers/iio/magnetometer/Makefile
> >> index 664b2f866472..df6ad23fee65 100644
> >> --- a/drivers/iio/magnetometer/Makefile
> >> +++ b/drivers/iio/magnetometer/Makefile
> >> @@ -24,3 +24,5 @@ obj-$(CONFIG_IIO_ST_MAGN_SPI_3AXIS) += st_magn_spi.o
> >>  obj-$(CONFIG_SENSORS_HMC5843)		+= hmc5843_core.o
> >>  obj-$(CONFIG_SENSORS_HMC5843_I2C)	+= hmc5843_i2c.o
> >>  obj-$(CONFIG_SENSORS_HMC5843_SPI)	+= hmc5843_spi.o
> >> +
> >> +obj-$(CONFIG_TLV493D) += tlv493d.o
> >> diff --git a/drivers/iio/magnetometer/tlv493d.c b/drivers/iio/magnetometer/tlv493d.c
> >> new file mode 100644
> >> index 000000000000..d2fe296b2c80
> >> --- /dev/null
> >> +++ b/drivers/iio/magnetometer/tlv493d.c
> >> @@ -0,0 +1,328 @@
> >> +// SPDX-License-Identifier: GPL-2.0+
> >> +/*
> >> + * Infineon TLV493D-A1B6 3D magnetic sensor  
> > 
> > link to datasheet would be nice  
> 
> https://www.infineon.com/cms/en/product/sensor/magnetic-position-sensor/3d-magnetic-sensor/tlv493d-a1b6/
> 
> Datasheet:
> https://www.infineon.com/dgdl/Infineon-TLV493D-A1B6-DS-v01_00-EN.pdf?fileId=5546d462525dbac40152a6b85c760e80
> 
> User Manual:
> https://www.infineon.com/dgdl/Infineon-TLV493D-A1B6_UM-UM-v01_02-EN.PDF?fileId=5546d462525dbac401530e81c9db058c
> 
> Not sure how stable the latter links are.
> 
> >   
> >> + *
> >> + * Copyright (c) 2016-2017 Andreas Färber
> >> + */
> >> +
> >> +#include <linux/bitfield.h>
> >> +#include <linux/i2c.h>
> >> +#include <linux/iio/iio.h>
> >> +#include <linux/module.h>
> >> +#include <linux/of.h>
> >> +#include <linux/of_device.h>
> >> +
> >> +#define MOD1_LOW	BIT(0)
> >> +#define MOD1_FAST	BIT(1)
> >> +#define MOD1_INT	BIT(2)
> >> +#define MOD1_P		BIT(7)
> >> +
> >> +#define MOD2_PT		BIT(5)  
> > 
> > please use TLV493D_ prefix;   
> 
> Okay.
> 
> > more descriptive naming or comments would be 
> > nice  
> 
> I'm sure those were the official field names from the manual.
> I can look into adding explanatory comments.
> 
> >> +struct tlv493d {
> >> +	struct i2c_client *i2c;
> >> +	struct iio_mount_matrix mount_matrix;
> >> +	u8 factset1, factset2, factset3;
> >> +};
> >> +
> >> +static int tlv493d_read_regs(struct tlv493d *s, u8 *buf, int len)
> >> +{
> >> +	int ret;
> >> +
> >> +	ret = i2c_master_recv(s->i2c, buf, len);
> >> +	if (ret < len) {
> >> +		dev_err(&s->i2c->dev, "failed to read registers (%d)", ret);
> >> +		return ret;
> >> +	}
> >> +
> >> +	return 0;
> >> +}
> >> +
> >> +static unsigned int tlv493d_parity(u32 v)
> >> +{
> >> +	v ^= v >> 16;
> >> +	v ^= v >> 8;
> >> +	v ^= v >> 4;
> >> +	v &= 0xf;
> >> +	return (0x6996 >> v) & 1;
> >> +}
> >> +
> >> +#define MOD1_RES_MASK		(0x3 << 3)
> >> +#define MOD1_IICADDR_MASK	(0x3 << 5)  
> > 
> > maybe move all these #defines on top of the file?  
> 
> Yeah, missed those while cleaning the code up.
> 
> >> +static int tlv493d_write_regs(struct tlv493d *s, const u8 *regs, int len)
> >> +{
> >> +	u8 buf[4];
> >> +	int ret;
> >> +
> >> +	if (len != ARRAY_SIZE(buf))
> >> +		return -EINVAL;  
> > 
> > why have this parameter if it is always 4?  
> 
> I originally had it as a BUG_ON() but checkpatch.pl complained about it.
> 
> const u8 *regs alone does not give us guarantees about its length, and
> to stay close to the manual I didn't want to mess with u32 and
> endianness. I'll have to re-check the manual where the limit stems from.

I think Peter was more meaning that we can do it by review.
This is fiddly but a static analyser such as smatch will probably
pick up any case where it was less than 4 and greater than 4 does
no harm (other than larger registers being ignored).
> 
> >> +	buf[0] = 0;
> >> +	buf[1] = s->factset1 & (MOD1_IICADDR_MASK | MOD1_RES_MASK);
> >> +	buf[1] |= regs[1] & ~(MOD1_P | MOD1_IICADDR_MASK | MOD1_RES_MASK);
> >> +	buf[2] = s->factset2;
> >> +	buf[3] = MOD2_PT | (s->factset3 & 0x1f);
> >> +	buf[3] |= regs[3] & ~(MOD2_PT | 0x1f);
> >> +
> >> +	if ((buf[3] & MOD2_PT) &&
> >> +	    !tlv493d_parity((buf[0] << 24) | (buf[1] << 16) | (buf[2] << 8) | buf[3]))
> >> +		buf[1] |= MOD1_P;
> >> +
> >> +	ret = i2c_master_send(s->i2c, buf, 4);  
> > 
> > use sizeof(buf) instead of 4?
> >   
> >> +	if (ret < 4) {
> >> +		dev_err(&s->i2c->dev, "failed to write registers (%d)", ret);
> >> +		return ret;
> >> +	}
> >> +
> >> +	return 0;
> >> +}
> >> +
> >> +static int tlv493d_power_down(struct tlv493d *s)
> >> +{
> >> +	u8 buf[4];  
> > 
> > maybe
> > u8 buf[4] = {0,};  
> >> +
> >> +	buf[0] = 0;
> >> +	buf[1] = 0;
> >> +	buf[2] = 0;
> >> +	buf[3] = 0;
> >> +	return tlv493d_write_regs(s, buf, 4);
> >> +}
> >> +
> >> +static const struct iio_mount_matrix *
> >> +tlv493d_get_mount_matrix(const struct iio_dev *indio_dev,
> >> +	const struct iio_chan_spec *chan)
> >> +{
> >> +	struct tlv493d *s = iio_priv(indio_dev);
> >> +
> >> +	return &s->mount_matrix;
> >> +}
> >> +
> >> +static const struct iio_chan_spec_ext_info tlv493d_ext_info[] = {
> >> +	IIO_MOUNT_MATRIX(IIO_SHARED_BY_DIR, tlv493d_get_mount_matrix),
> >> +	{}
> >> +};
> >> +
> >> +#define TLV493D_AXIS_CHANNEL(axis, index)			\
> >> +	{							\
> >> +		.type = IIO_MAGN,				\
> >> +		.channel2 = IIO_MOD_##axis,			\
> >> +		.modified = 1,					\
> >> +		.address = index,				\
> >> +		.scan_index = index,				\
> >> +		.scan_type = {					\  
> > 
> > IIO buffer reads are not supported by the driver, so no need for 
> > .scan_type (here and for _TEMP)  
> 
> Hmm, I wanted to get the code out before it catches too much dust here,
> I might revisit IIO buffers later.
> 
> The general problem I've run into is that the values vary quite a bit
> between readings, and the iio API seemed to force me into doing I2C
> readings per channel. Do you see a better way (any example?) to read the
> three channels once and then return the three axes from one consistent
> reading?

Use the buffered interface.  It's that simple - it is near impossible
to define a generic interface for getting a consistent set of values
across an unknown number of channels in sysfs.  Unknown in the sense
that the the ABI definition doesn't know - obvious a given driver
does.

Sysfs is one value one attribute.  There are odd well defined
multi element things such as quaternions but in those cases they really
aren't separable - with axis measurements they are.

> 
> >> +			.sign = 's',				\
> >> +			.realbits = 12,				\
> >> +			.storagebits = 16,			\
> >> +			.endianness = IIO_CPU,			\
> >> +		},						\
> >> +		.info_mask_separate = BIT(IIO_CHAN_INFO_RAW) |	\
> >> +			BIT(IIO_CHAN_INFO_PROCESSED),		\  
> > 
> > either _RAW or _PROCESSED, not both  
> 
> Can you explain please? I tested this via sysfs and tools and could
> access both.
Indeed. It's possible of course, however the ABI is not supposed
to be redundant.  There are a few exceptions where we had existing
ABI due to say supporting raw only, but later it became apparent
a non linear conversion was needed so we ended up with both.

Given IIO also supports some high speed devices where in kernel conversion
is too costly, and others where is in accurate due to not having floating
point maths - the decision from the start years ago was to default to
_raw unless there was a reason we couldn't - typically non linear
conversions on light sensors.

> 
> >> +		.info_mask_shared_by_type = BIT(IIO_CHAN_INFO_SCALE),	\
> >> +		.ext_info = tlv493d_ext_info,			\
> >> +	}
> >> +
> >> +#define TLV493D_TEMP_CHANNEL(index)				\
> >> +	{							\
> >> +		.type = IIO_TEMP,				\
> >> +		.channel2 = IIO_MOD_TEMP_OBJECT,		\
> >> +		.modified = 1,					\
> >> +		.address = index,				\
> >> +		.scan_index = index,				\
> >> +		.scan_type = {					\
> >> +			.sign = 's',				\
> >> +			.realbits = 12,				\
> >> +			.storagebits = 16,			\
> >> +			.endianness = IIO_CPU,			\
> >> +		},						\
> >> +		.info_mask_separate = BIT(IIO_CHAN_INFO_RAW) |	\
> >> +			BIT(IIO_CHAN_INFO_OFFSET) |		\
> >> +			BIT(IIO_CHAN_INFO_SCALE) |		\
> >> +			BIT(IIO_CHAN_INFO_PROCESSED),		\
> >> +	}
> >> +
> >> +static const struct iio_chan_spec tlv493d_channels[] = {
> >> +	TLV493D_AXIS_CHANNEL(X, 0),
> >> +	TLV493D_AXIS_CHANNEL(Y, 1),
> >> +	TLV493D_AXIS_CHANNEL(Z, 2),
> >> +	TLV493D_TEMP_CHANNEL(3),
> >> +	IIO_CHAN_SOFT_TIMESTAMP(4),
> >> +};
> >> +
> >> +static const unsigned long tlv493d_scan_masks[] = { 0xf, 0 };
> >> +
> >> +static int tlv493d_read_raw(struct iio_dev *indio_dev,
> >> +	struct iio_chan_spec const *chan,
> >> +	int *val, int *val2, long mask)
> >> +{
> >> +	struct tlv493d *s = iio_priv(indio_dev);
> >> +	u8 buf[7];
> >> +	int ret;
> >> +
> >> +	switch (mask) {
> >> +	case IIO_CHAN_INFO_PROCESSED:
> >> +	case IIO_CHAN_INFO_RAW:  
> > 
> > this is ugly, why support _PROCESSED? 
> > just _RAW is fine  
> 
> See below...
> 
> >   
> >> +		do {
> >> +			ret = tlv493d_read_regs(s, buf, 7);
> >> +			if (ret)
> >> +				return ret;
> >> +		} while ((buf[3] & 0x3) || (buf[5] & BIT(4)));  
> > 
> > maybe add a timeout or limit the number of retries
> >   
> >> +
> >> +		switch (chan->address) {
> >> +		case 0:
> >> +			*val = (((int)(s8)buf[0]) << 4) | (buf[4] >> 4);
> >> +			break;
> >> +		case 1:
> >> +			*val = (((int)(s8)buf[1]) << 4) | (buf[4] & 0xf);
> >> +			break;
> >> +		case 2:
> >> +			*val = (((int)(s8)buf[2]) << 4) | (buf[5] & 0xf);
> >> +			break;
> >> +		case 3:
> >> +			*val = (((int)(s8)(buf[3] & 0xf0)) << 4) | buf[6];
> >> +			if (mask != IIO_CHAN_INFO_RAW)
> >> +				*val -= 340;  
> 
> Here an offset is being applied to turn the temp value into something
> real. More differences below.

Then represent it in _offset for the channel.

> 
> >> +			break;
> >> +		default:
> >> +			return -EINVAL;
> >> +		}
> >> +		*val2 = 0;  
> > 
> > no need to set val2 if returning VAL_INT  
> 
> Okay.
> 
> >> +		if (mask == IIO_CHAN_INFO_RAW)
> >> +			return IIO_VAL_INT;  
> 
> Here the RAW value is returned, whereas some more corrections are
> applied to the values below. The raw value was mostly for debugging or
> if someone wants to apply different correction factors.

The below can defintiely be represented as scale and offset so just
do that and stick to raw.

> 
> >> +
> >> +		switch (chan->type) {
> >> +		case IIO_MAGN:
> >> +			*val2 = (*val * 1000000) * 98 / 1000;
> >> +			*val = *val2 / 1000000;
> >> +			*val2 %= 1000000;
> >> +			break;
> >> +		case IIO_TEMP:
> >> +			*val2 = (*val * 1000000) * 11 / 10;
> >> +			*val = *val2 / 1000000;
> >> +			*val2 %= 1000000;
> >> +			/* According to datasheet, LSB offset is for 25°C */
> >> +			*val += 25;
> >> +			break;
> >> +		default:
> >> +			return -EINVAL;
> >> +		}
> >> +		return IIO_VAL_INT_PLUS_MICRO;
> >> +	case IIO_CHAN_INFO_OFFSET:
> >> +		switch (chan->type) {
> >> +		case IIO_TEMP:
> >> +			*val = -340;
> >> +			*val2 = 0;  
> > 
> > no need to set val2  
> 
> Okay.
> 
> >> +			return IIO_VAL_INT;
> >> +		default:
> >> +			return -EINVAL;
> >> +		}
> >> +	case IIO_CHAN_INFO_SCALE:
> >> +		switch (chan->type) {
> >> +		case IIO_MAGN:
> >> +			/* 0.098 mT/LSB */
> >> +			*val = 0;
> >> +			*val2 = 98000;
> >> +			return IIO_VAL_INT_PLUS_MICRO;
> >> +		case IIO_TEMP:
> >> +			/* 1.1 °C/LSB */
> >> +			*val = 1;
> >> +			*val2 = 100000;
> >> +			return IIO_VAL_INT_PLUS_MICRO;
> >> +		default:
> >> +			return -EINVAL;
> >> +		}
> >> +	default:
> >> +		return -EINVAL;
> >> +	}
> >> +}
> >> +
> >> +static const struct iio_info tlv493d_iio_info = {
> >> +	.read_raw = tlv493d_read_raw,
> >> +};
> >> +
> >> +static int tlv493d_probe(struct i2c_client *i2c)
> >> +{
> >> +	struct iio_dev *indio_dev;
> >> +	struct tlv493d *tlv493d;
> >> +	u8 buf[10];
> >> +	int ret;
> >> +
> >> +	indio_dev = devm_iio_device_alloc(&i2c->dev, sizeof(*tlv493d));
> >> +	if (!indio_dev)
> >> +		return -ENOMEM;
> >> +
> >> +	tlv493d = iio_priv(indio_dev);
> >> +	i2c_set_clientdata(i2c, indio_dev);
> >> +	tlv493d->i2c = i2c;
> >> +
> >> +	ret = of_iio_read_mount_matrix(&i2c->dev, "mount-matrix",
> >> +		&tlv493d->mount_matrix);
> >> +	if (ret)
> >> +		return ret;
> >> +
> >> +	indio_dev->dev.parent = &i2c->dev;
> >> +	indio_dev->channels = tlv493d_channels;
> >> +	indio_dev->num_channels = ARRAY_SIZE(tlv493d_channels);
> >> +	indio_dev->info = &tlv493d_iio_info;
> >> +	indio_dev->available_scan_masks = tlv493d_scan_masks;
> >> +	indio_dev->modes = INDIO_DIRECT_MODE;
> >> +	indio_dev->name = "tlv493d";
> >> +
> >> +	ret = tlv493d_read_regs(tlv493d, buf, 10);
> >> +	if (ret)
> >> +		return ret;
> >> +
> >> +	tlv493d->factset1 = buf[7];
> >> +	tlv493d->factset2 = buf[8];
> >> +	tlv493d->factset3 = buf[9];
> >> +
> >> +	buf[0] = 0;
> >> +	buf[1] = MOD1_FAST | MOD1_LOW;  
> > 
> > what does this mean? some comment explaining the default mode?  
> 
> AFAIR the chip can operate in some non-standard-I2C modes (~four), but
> it's been some time since I wrote this... I'll dig into it.
> 
> >> +	buf[2] = 0;
> >> +	buf[3] = 0;
> >> +	ret = tlv493d_write_regs(tlv493d, buf, 4);
> >> +	if (ret)
> >> +		return ret;
> >> +
> >> +	ret = iio_device_register(indio_dev);
> >> +	if (ret) {
> >> +		dev_err(&i2c->dev, "device registration failed");
> >> +		tlv493d_power_down(tlv493d);
> >> +		return ret;
> >> +	}
> >> +
> >> +	return 0;
> >> +}
> >> +
> >> +static int tlv493d_remove(struct i2c_client *i2c)
> >> +{
> >> +	struct iio_dev *indio_dev = i2c_get_clientdata(i2c);
> >> +	struct tlv493d *tlv493d = iio_priv(indio_dev);
> >> +
> >> +	iio_device_unregister(indio_dev);
> >> +
> >> +	tlv493d_power_down(tlv493d);
> >> +
> >> +	return 0;
> >> +}
> >> +
> >> +static const struct of_device_id tlv493d_dt_ids[] = {
> >> +	{ .compatible = "infineon,tlv493d-a1b6" },
> >> +	{}
> >> +};
> >> +MODULE_DEVICE_TABLE(of, tlv493d_dt_ids);
> >> +
> >> +static struct i2c_driver tlv493d_i2c_driver = {
> >> +	.driver = {
> >> +		.name = "tlv493d",
> >> +		.of_match_table = tlv493d_dt_ids,
> >> +	},
> >> +	.probe_new = tlv493d_probe,
> >> +	.remove = tlv493d_remove,
> >> +};
> >> +
> >> +module_i2c_driver(tlv493d_i2c_driver);
> >> +
> >> +MODULE_DESCRIPTION("TLV493D I2C driver");
> >> +MODULE_AUTHOR("Andreas Faerber");
> >> +MODULE_LICENSE("GPL");  
> 
> Thanks for your quick review,
> 
> Andreas
> 

Powered by blists - more mailing lists