Message-ID: <548C209A.7000904@gmx.de>
Date:	Sat, 13 Dec 2014 12:18:50 +0100
From:	Hartmut Knaack <knaack.h@....de>
To:	Nikolaus Schulz <nikolaus.schulz@...onic-design.de>
CC:	Jonathan Cameron <jic23@...nel.org>,
	Lars-Peter Clausen <lars@...afoo.de>,
	Peter Meerwald <pmeerw@...erw.net>,
	Grant Likely <grant.likely@...aro.org>,
	Rob Herring <robh+dt@...nel.org>,
	Michael Welling <mwelling@...e.org>,
	Philippe Reynes <tremyfr@...oo.fr>,
	linux-kernel@...r.kernel.org, linux-iio@...r.kernel.org,
	devicetree@...r.kernel.org,
	Alban Bedel <alban.bedel@...onic-design.de>
Subject: Re: [PATCH v2 2/2] iio: add driver for the TI DAC8554

Nikolaus Schulz schrieb am 12.12.2014 um 16:58:
> On Sat, Dec 06, 2014 at 12:36:19PM +0100, Hartmut Knaack wrote:
>> Nikolaus Schulz schrieb am 24.11.2014 um 20:50:
>>> The TI DAC8554 is a quad-channel Digital-to-Analog Converter with an SPI
>>> interface.
>>>
>>> Changes in v2:
>>> * Use DMA-safe buffer for SPI transfer
>>> * Normalize powerdown_mode name "hi-z" to "three_state" as per
>>>   ABI/testing/sysfs-bus-iio
>>> * Register device late in probe function
>>> * Avoid powerdown broadcast update, which touches all DAC on the bus
>> There are a few issues left, please see my comments inline.
>>>
>>> Signed-off-by: Nikolaus Schulz <nikolaus.schulz@...onic-design.de>
>>> ---
>>>  drivers/iio/dac/Kconfig      |  10 ++
>>>  drivers/iio/dac/Makefile     |   1 +
>>>  drivers/iio/dac/ti-dac8554.c | 374 +++++++++++++++++++++++++++++++++++++++++++
>>>  3 files changed, 385 insertions(+)
>>>  create mode 100644 drivers/iio/dac/ti-dac8554.c
> [...]
>>> diff --git a/drivers/iio/dac/ti-dac8554.c b/drivers/iio/dac/ti-dac8554.c
>>> new file mode 100644
>>> index 0000000..fca751f
>>> --- /dev/null
>>> +++ b/drivers/iio/dac/ti-dac8554.c
>>> @@ -0,0 +1,374 @@
>>> +/*
>>> + * TI DAC8554 Digital to Analog Converter SPI driver
>>> + *
>>> + * Copyright (C) 2014 Avionic Design GmbH
>>> + *
>>> + * Based on ad5446r_spi.c
>>> + * Copyright (C) 2010,2011 Analog Devices Inc.
>>> + *
>>> + * This program is free software; you can redistribute it and/or modify
>>> + * it under the terms of the GNU General Public License version 2 as
>>> + * published by the Free Software Foundation.
>>> + */
>>> +
>>> +#include <linux/spi/spi.h>
>>> +#include <linux/regulator/consumer.h>
>>> +#include <linux/module.h>
>>> +#include <linux/iio/iio.h>
>>> +#include <linux/of.h>
>>> +
>>> +#define DAC8554_DRIVERNAME			"ti-dac8554"
>>> +#define DAC8554_DAC_CHANNELS			4
>>> +
>>> +/* Load commands */
>>> +#define DAC8554_CMD_STORE_DAC_N			0x0
>>> +#define DAC8554_CMD_UPDATE_DAC_N		0x1
>>> +#define DAC8554_CMD_STORE_DAC_N_UPDATE_ALL	0x2
>>> +#define DAC8554_CMD_UPDATE_BROADCAST		0x3
>>> +
>>> +#define DAC8554_BROADCAST_USE_SRDATA		0x2
>> Add a blank line here.
> 
> Fixed.
> 
>>> +/* Powerdown modes (ORed PD1|PD2 bits) */
>> It took my some time to figure out ORed. I think you could drop it and just
>> mention (PD1 | PD2 bits).
> 
> Agreed, and fixed.
> 
>>> +#define DAC8554_PWRDN_HIZ			0x0
>>> +#define DAC8554_PWRDN_1K			0x1
>>> +#define DAC8554_PWRDN_100K			0x2
>>> +
>>> +#define DAC8554_PWRDN_TO_SR(mode)		(BIT(16) | (mode) << 14)
>>> +
>>> +/**
>>> + * struct dac8554_state - driver instance specific data
>>> + * @spi:		spi_device
>> SPI device?
> 
> Yes, fixed.
> 
>>> + * @reg:		supply regulator
>>> + * @addr:		two-bit chip address
>>> + * @vref_mv:		reference voltage in millivolt
>>> + * @val			DAC/channel data
>> DAC channel data?
> 
> "channel data" will do, actually.
> 
>>> + * @powerdown		channel powerdown flag
>>> + * @powerdown_mode	channel powerdown mode
>>> + * @xfer		SPI transfer buffer
>> Half of the comments are with colon, the other half are missing it.
> 
> Fixed.
> 
>>> + */
>>> +struct dac8554_state {
>>> +	struct spi_device		*spi;
>>> +	struct regulator		*reg;
>>> +	unsigned			addr;
>>> +	unsigned			vref_mv;
>>> +	u16				val[DAC8554_DAC_CHANNELS];
>>> +	bool				powerdown[DAC8554_DAC_CHANNELS];
>>> +	u8				powerdown_mode[DAC8554_DAC_CHANNELS];
>>> +
>>> +	/*
>>> +	 * DMA (thus cache coherency maintenance) requires the
>>> +	 * transfer buffers to live in their own cache lines.
>>> +	 */
>>> +	u8				xfer[3] ____cacheline_aligned;
>>> +};
>>> +
>>> +static int dac8554_spi_write(struct dac8554_state *st,
>>> +			     unsigned cmd,
>>> +			     unsigned chan_addr,
>>> +			     unsigned val)
>>> +{
>>> +	u32 data;
>>> +
>>> +	/*
>>> +	 * The input shift register is 24 bits wide. The 8 MSB are
>>> +	 * control bits, followed by 16 data bits.
>>> +	 * The first two bits A1 and A0 address a DAC8554 chip.
>>> +	 * The next two are the command bits, LD1 and LD0.
>>> +	 * After a don't-care-bit, the next two bits select the channel.
>>> +	 * The final control bit PD0 is a flag signalling if the data
>>> +	 * bits encode a powerdown mode. We merge PD0 with the adjacent
>>> +	 * data bits.
>>> +	 */
>>> +
>>> +	if (cmd > 3 || chan_addr > 3 ||
>>> +			(val > 0xffff && (val & ~DAC8554_PWRDN_TO_SR(3))))
>>> +		return -EINVAL;
>>> +
>>> +	data =st->addr << 22) | (cmd << 20) | (chan_addr << 17) | val;
>> Could be more readable by using definition for the magic shifts, or macros.
> 
> Hmm. The magic is all concentrated at this place, and the variable names
> tell the story, so I figured that's ok.  But, why not: fixed.
> 
>>> +	st->xfer[0] =ata >> 16;
>>> +	st->xfer[1] =ata >> 8;
>>> +	st->xfer[2] =ata;
>>> +
>>> +	return spi_write(st->spi, st->xfer, sizeof(st->xfer));
>>> +}
>>> +
>>> +static int dac8554_read_raw(struct iio_dev *indio_dev,
>>> +			    struct iio_chan_spec const *chan,
>>> +			    int *val,
>>> +			    int *val2,
>>> +			    long m)
>> Commonly, m is called mask (like you do in _write_raw).
> 
> Ah yes. Fixed.
> 
>>> +{
>>> +	struct dac8554_state *st =io_priv(indio_dev);
>>> +
>>> +	switch (m) {
>>> +	case IIO_CHAN_INFO_RAW:
>>> +		*val =t->val[chan->address];
>>> +		return IIO_VAL_INT;
>>> +	case IIO_CHAN_INFO_SCALE:
>>> +		*val =t->vref_mv;
>>> +		*val2 =6;
>>> +		return IIO_VAL_FRACTIONAL_LOG2;
>>> +	}
>>> +	return -EINVAL;
>>> +}
>>> +
>>> +static int dac8554_write_raw(struct iio_dev *indio_dev,
>>> +			     struct iio_chan_spec const *chan,
>>> +			     int val,
>>> +			     int val2,
>>> +			     long mask)
>>> +{
>>> +	struct dac8554_state *st =io_priv(indio_dev);
>>> +	int err;
>>> +
>>> +	switch (mask) {
>>> +	case IIO_CHAN_INFO_RAW:
>>> +		if (val > 0xffff || val < 0)
>>> +			return -EINVAL;
>>> +
>>> +		err =ac8554_spi_write(st, DAC8554_CMD_UPDATE_DAC_N,
>>> +					chan->address, val);
>>> +		if (err)
>>> +			return err;
>>> +
>>> +		st->val[chan->address] =al;
>>> +
>>> +		/* By hw design, DAC updates automatically trigger powerup. */
>>> +		st->powerdown[chan->address] =alse;
>>> +
>>> +		return 0;
>>> +
>>> +	default:
>>> +		return -EINVAL;
>>> +	}
>>> +}
>>> +
>>> +static int dac8554_get_powerdown_mode(struct iio_dev *indio_dev,
>>> +				      const struct iio_chan_spec *chan)
>>> +{
>>> +	struct dac8554_state *st =io_priv(indio_dev);
>>> +
>>> +	return st->powerdown_mode[chan->address];
>>> +}
>>> +
>>> +static int dac8554_set_powerdown_mode(struct iio_dev *indio_dev,
>>> +				      const struct iio_chan_spec *chan,
>>> +				      unsigned int mode)
>>> +{
>>> +	struct dac8554_state *st =io_priv(indio_dev);
>>> +
>>> +	st->powerdown_mode[chan->address] =ode;
>>> +
>>> +	return 0;
>>> +}
>>> +
>>> +static ssize_t dac8554_read_dac_powerdown(struct iio_dev *indio_dev,
>>> +					  uintptr_t private,
>>> +					  const struct iio_chan_spec *chan,
>>> +					  char *buf)
>>> +{
>>> +	struct dac8554_state *st =io_priv(indio_dev);
>>> +
>>> +	return sprintf(buf, "%d\n", st->powerdown[chan->address]);
>>> +}
>>> +
>>> +static ssize_t dac8554_write_dac_powerdown(struct iio_dev *indio_dev,
>>> +					   uintptr_t private,
>>> +					   const struct iio_chan_spec *chan,
>>> +					   const char *buf,
>>> +					   size_t len)
>>> +{
>>> +	bool powerdown;
>>> +	int ret;
>>> +	struct dac8554_state *st =io_priv(indio_dev);
>>> +	u8 powerdown_mode;
>>> +
>>> +	ret =trtobool(buf, &powerdown);
>>> +	if (ret)
>>> +		return ret;
>>> +
>>> +	st->powerdown[chan->address] =owerdown;
>>> +
>>> +	if (powerdown) {
>>> +		powerdown_mode =t->powerdown_mode[chan->address];
>>> +		ret =ac8554_spi_write(st,
>>> +				DAC8554_CMD_UPDATE_DAC_N,
>>> +				chan->address,
>>> +				DAC8554_PWRDN_TO_SR(powerdown_mode));
>> The parameters sill fit on line even with proper indentation.
> 
> Fixed.
> 
>>> +	} else {
>>> +		/* Load DAC with cached value. This triggers a powerup. */
>>> +		ret =ac8554_spi_write(st,
>>> +				DAC8554_CMD_UPDATE_DAC_N,
>>> +				chan->address,
>>> +				st->val[chan->address]);
>> Same here with indentation.
> 
> Fixed.
> 
>>> +	}
>>> +
>>> +	if (ret)
>>> +		return ret;
>>> +
>>> +	return len;
>> Just: return (ret) ? ret : len;
> 
> I think avoiding the ternary operator makes the code more readable, and
> it separates the error path from normal operation. So I would prefer to
> keep it that way.
Fair enough.
> 
>>> +}
>>> +
>>> +static int dac8554_powerdown(struct dac8554_state *st,
>>> +			     u8 powerdown_mode)
>>> +{
>>> +	int chan, cmd, ret;
>>> +
>>> +	for (chan =AC8554_DAC_CHANNELS-1; chan >= 0; --chan) {
>> Please mind spaces around operators.
> 
> Fixed.
> 
>>> +		cmd =han ? DAC8554_CMD_STORE_DAC_N
>>> +			   : DAC8554_CMD_STORE_DAC_N_UPDATE_ALL;
>>> +		ret =ac8554_spi_write(st, cmd, chan,
>>> +				DAC8554_PWRDN_TO_SR(powerdown_mode));
>>> +		if (ret)
>>> +			return ret;
>>> +	}
>>> +
>>> +	memset(st->powerdown_mode, powerdown_mode, sizeof(st->powerdown_mode));
>>> +	memset(st->powerdown, true, sizeof(st->powerdown));
>>> +
>>> +	return 0;
>>> +}
>>> +
>>> +static const struct iio_info dac8554_info =
>>> +	.write_raw = ac8554_write_raw,
>>> +	.read_raw =ac8554_read_raw,
>>> +	.driver_module =HIS_MODULE,
>>> +};
>>> +
>>> +static const char * const dac8554_powerdown_modes[] =
>>> +	"three_state",
>>> +	"1kohm_to_gnd",
>>> +	"100kohm_to_gnd"
>>> +};
>>> +
>>> +static const struct iio_enum dac8854_powerdown_mode_enum =
>>> +	.items = ac8554_powerdown_modes,
>>> +	.num_items =RRAY_SIZE(dac8554_powerdown_modes),
>>> +	.get =ac8554_get_powerdown_mode,
>>> +	.set =ac8554_set_powerdown_mode,
>>> +};
>>> +
>>> +static const struct iio_chan_spec_ext_info dac8554_ext_info[] =
>>> +	{
>>> +		.name =powerdown",
>>> +		.read =ac8554_read_dac_powerdown,
>>> +		.write =ac8554_write_dac_powerdown,
>>> +		.shared =IO_SEPARATE,
>>> +	},
>>> +	IIO_ENUM("powerdown_mode", IIO_SEPARATE,
>>> +		 &dac8854_powerdown_mode_enum),
>>> +	IIO_ENUM_AVAILABLE("powerdown_mode", &dac8854_powerdown_mode_enum),
>>> +	{ },
>>> +};
>>> +
>>> +#define DAC8554_CHANNEL(chan) { \
>>> +	.type =IO_VOLTAGE, \
>>> +	.indexed =, \
>>> +	.output =, \
>>> +	.channel =chan), \
>>> +	.info_mask_separate =IT(IIO_CHAN_INFO_RAW), \
>>> +	.info_mask_shared_by_type =IT(IIO_CHAN_INFO_SCALE), \
>>> +	.address =chan), \
>>> +	.ext_info =ac8554_ext_info, \
>>> +}
>>> +const struct iio_chan_spec dac8554_channels[] =
>>> +	DAC8554_CHANNEL(0),
>>> +	DAC8554_CHANNEL(1),
>>> +	DAC8554_CHANNEL(2),
>>> +	DAC8554_CHANNEL(3),
>>> +};
>>> +#undef DAC8554_CHANNEL
>>> +
>>> +static int dac8554_probe(struct spi_device *spi)
>>> +{
>>> +	struct dac8554_state *st;
>>> +	struct iio_dev *indio_dev;
>>> +	int ret, voltage_uv;
>>> +	u32 addr;
>>> +
>>> +	indio_dev =evm_iio_device_alloc(&spi->dev, sizeof(*st));
>>> +	if (!indio_dev)
>>> +		return -ENOMEM;
>>> +
>>> +	indio_dev->dev.parent =spi->dev;
>>> +	indio_dev->name =AC8554_DRIVERNAME;
>>> +	indio_dev->info =dac8554_info;
>>> +	indio_dev->modes =NDIO_DIRECT_MODE;
>>> +	indio_dev->channels =ac8554_channels;
>>> +	indio_dev->num_channels =RRAY_SIZE(dac8554_channels);
>>> +
>>> +	spi_set_drvdata(spi, indio_dev);
>>> +
>>> +	st =io_priv(indio_dev);
>>> +
>>> +	if (!spi->dev.of_node) {
>>> +		dev_err(&spi->dev, "missing OF node");
>>> +		return -ENODEV;
>>> +	}
>>> +	ret =f_property_read_u32(spi->dev.of_node, "address", &addr);
>>> +	if (ret || addr < 0 || addr > 2) {
>>> +		dev_err(&spi->dev, "no or invalid chip address");
>>> +		return -ENODEV;
>>> +	}
>>> +
>>> +	st->spi =pi;
>>> +	st->addr =ddr;
>>> +
>> According to your DT bindings, the regulator from property "vref-supply" should
>> be used. This is missing here.
> 
> Uhm, it's right below, no?
Looking into your DT bindings patch (which unfortunately didn't make it into our
list), you specify "vref-supply" as a required property. So, I expected that you
intended to make use of this property somewhere around here.
> 
>>> +	st->reg =evm_regulator_get(&spi->dev, "vref");
>>> +	if (IS_ERR(st->reg))
>>> +		return PTR_ERR(st->reg);
>>> +
>>> +	ret =egulator_enable(st->reg);
>>> +	if (ret)
>>> +		return ret;
>>> +
>>> +	voltage_uv =egulator_get_voltage(st->reg);
>>> +	if (voltage_uv < 0)
>>> +		goto error_disable_reg;
>> Missing ret =oltage_uv before goto. Or just drop voltage_uv completely and use ret instead.
> 
> Fixed.
> 
>>> +	st->vref_mv =oltage_uv / 1000;
>> How hard do you want to depend on a voltage regulator? Doing regulator_get_voltage()
>> could even be called dynamically in _read_raw(), making a real regulator optional.
> 
> Hmm. I understand that the DAC voltage input may not be provided by a
> regulator, but is that a common scenario?  No other DAC driver I looked
> at handles that case, they all consider it an error if the regulator is
> absent.
> 
I had people in mind, who might want to use such DAC with one of the popular
embedded boards like Raspberry Pi - people who would just like to "plug and
play" without the need or ability to build a kernel (or would prefer to go
with the convenience of a distribution kernel). But since you've put a
dependency on OF, I don't see a way to use this device with a stock distribution
kernel anyway.
But I still think that calculating the actual voltage during _read_raw() would
be a good idea, as there is support for regulators providing different voltages,
which can be changed during runtime.
>>> +
>>> +	ret =ac8554_powerdown(st, DAC8554_PWRDN_100K);
>>> +	if (ret)
>>> +		goto error_disable_reg;
>>> +
>>> +	ret =io_device_register(indio_dev);
>>> +	if (ret)
>>> +		goto error_disable_reg;
>>> +
>>> +	return 0;
>>> +
>>> +error_disable_reg:
>>> +	regulator_disable(st->reg);
>>> +	return ret;
>>> +}
>>> +
>>> +static int dac8554_remove(struct spi_device *spi)
>>> +{
>>> +	struct iio_dev *indio_dev =pi_get_drvdata(spi);
>>> +	struct dac8554_state *st =io_priv(indio_dev);
>>> +
>>> +	iio_device_unregister(indio_dev);
>>> +	regulator_disable(st->reg);
>>> +
>>> +	return 0;
>>> +}
>>> +
>>> +static const struct of_device_id dac8554_of_match[] =
>>> +	{ .compatible = ti,dac8554" },
>>> +	{ }
>>> +};
>>> +MODULE_DEVICE_TABLE(of, dac8554_of_match);
>>> +
>>> +static struct spi_driver dac8554_driver =
>>> +	.driver = 
>>> +		.name = AC8554_DRIVERNAME,
>>> +		.owner =HIS_MODULE,
>>> +		.of_match_table =ac8554_of_match,
>>> +	 },
>>> +	.probe =ac8554_probe,
>>> +	.remove =ac8554_remove,
>>> +};
>>> +module_spi_driver(dac8554_driver);
>>> +
>>> +MODULE_AUTHOR("Nikolaus Schulz <nikolaus.schulz@...onic-design.de>");
>>> +MODULE_DESCRIPTION("Texas Instruments DAC8554 SPI driver");
>>> +MODULE_LICENSE("GPL v2");
> 
> Thanks for the review!
> 
> Nikolaus
> 

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

Powered by blists - more mailing lists