On Sun, 2017-04-23 at 20:13 +0100, Jonathan Cameron wrote:
> On 21/04/17 21:19, Peter Meerwald-Stadler wrote:
> >
> >> From: Mårten Lindahl <martenli@xxxxxxxx>
> >
> > comments below
> A few more from me. Laptop battery gone flat so not so thorough on top few lines!
>
> Jonathan
Hi, and thanks for your comments! I have a new patch where I resolved it
all. But before I send it I wonder about your comments about the events.
Please see below.
Thanks,
Mårten
> >
> >> This adds support for the Texas Instruments ADC084S021 ADC chip.
> >>
> >> Signed-off-by: Mårten Lindahl <martenli@xxxxxxxx>
> >> ---
> >> .../devicetree/bindings/iio/adc/ti-adc084s021.txt | 25 ++
> >> drivers/iio/adc/Kconfig | 12 +
> >> drivers/iio/adc/Makefile | 1 +
> >> drivers/iio/adc/ti-adc084s021.c | 342 +++++++++++++++++++++
> >> 4 files changed, 380 insertions(+)
> >> create mode 100644 Documentation/devicetree/bindings/iio/adc/ti-adc084s021.txt
> >> create mode 100644 drivers/iio/adc/ti-adc084s021.c
> >>
> >> diff --git a/Documentation/devicetree/bindings/iio/adc/ti-adc084s021.txt b/Documentation/devicetree/bindings/iio/adc/ti-adc084s021.txt
> >> new file mode 100644
> >> index 0000000..921eb46
> >> --- /dev/null
> >> +++ b/Documentation/devicetree/bindings/iio/adc/ti-adc084s021.txt
> >> @@ -0,0 +1,25 @@
> >> +* Texas Instruments' ADC084S021
> >> +
> >> +Required properties:
> >> + - compatible : Must be "ti,adc084s021"
> >> + - reg : SPI chip select number for the device
> >> + - vref-supply : The regulator supply for ADC reference voltage
> >> + - spi-max-frequency : Definition as per Documentation/devicetree/bindings/spi/spi-bus.txt
> >> +
> >> +Optional properties:
> >> + - spi-cpol : SPI inverse clock polarity, as per spi-bus bindings
> >> + - spi-cpha : SPI shifted clock phase (CPHA), as per spi-bus bindings
> >> + - spi-cs-high : SPI chip select active high, as per spi-bus bindings
> >> +
> >> +
> >> +Example:
> >> +adc@0 {
> >> + compatible = "ti,adc084s021";
> >> + reg = <0>;
> >> + vref-supply = <&adc_vref>;
> >> + spi-cpol;
> >> + spi-cpha;
> >> + spi-cs-high;
> >> + spi-max-frequency = <16000000>;
> >> + pl022,com-mode = <0x2>; /* DMA */
> >
> > what is this?
> >
> >> +};
> >> diff --git a/drivers/iio/adc/Kconfig b/drivers/iio/adc/Kconfig
> >> index dedae7a..13141e5 100644
> >> --- a/drivers/iio/adc/Kconfig
> >> +++ b/drivers/iio/adc/Kconfig
> >> @@ -560,6 +560,18 @@ config TI_ADC0832
> >> This driver can also be built as a module. If so, the module will be
> >> called ti-adc0832.
> >>
> >> +config TI_ADC084S021
> >> + tristate "Texas Instruments ADC084S021"
> >> + depends on SPI
> >> + select IIO_BUFFER
> >> + select IIO_TRIGGERED_BUFFER
> >> + help
> >> + If you say yes here you get support for Texas Instruments ADC084S021
> >> + chips.
> >> +
> >> + This driver can also be built as a module. If so, the module will be
> >> + called ti-adc084s021.
> >> +
> >> config TI_ADC12138
> >> tristate "Texas Instruments ADC12130/ADC12132/ADC12138"
> >> depends on SPI
> >> diff --git a/drivers/iio/adc/Makefile b/drivers/iio/adc/Makefile
> >> index d001262..b1a6158 100644
> >> --- a/drivers/iio/adc/Makefile
> >> +++ b/drivers/iio/adc/Makefile
> >> @@ -51,6 +51,7 @@ obj-$(CONFIG_STM32_ADC_CORE) += stm32-adc-core.o
> >> obj-$(CONFIG_STM32_ADC) += stm32-adc.o
> >> obj-$(CONFIG_TI_ADC081C) += ti-adc081c.o
> >> obj-$(CONFIG_TI_ADC0832) += ti-adc0832.o
> >> +obj-$(CONFIG_TI_ADC084S021) += ti-adc084s021.o
> >> obj-$(CONFIG_TI_ADC12138) += ti-adc12138.o
> >> obj-$(CONFIG_TI_ADC128S052) += ti-adc128s052.o
> >> obj-$(CONFIG_TI_ADC161S626) += ti-adc161s626.o
> >> diff --git a/drivers/iio/adc/ti-adc084s021.c b/drivers/iio/adc/ti-adc084s021.c
> >> new file mode 100644
> >> index 0000000..4f33b91
> >> --- /dev/null
> >> +++ b/drivers/iio/adc/ti-adc084s021.c
> >> @@ -0,0 +1,342 @@
> >> +/**
> >> + * Copyright (C) 2017 Axis Communications AB
> >> + *
> >> + * Driver for Texas Instruments' ADC084S021 ADC chip.
> >> + * Datasheets can be found here:
> >> + * http://www.ti.com/lit/ds/symlink/adc084s021.pdf
> >> + *
> >> + * 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/err.h>
> >> +#include <linux/spi/spi.h>
> >> +#include <linux/module.h>
> >> +#include <linux/interrupt.h>
> >> +#include <linux/iio/iio.h>
> >> +#include <linux/iio/buffer.h>
> >> +#include <linux/iio/events.h>
> >> +#include <linux/iio/triggered_buffer.h>
> >> +#include <linux/iio/trigger_consumer.h>
> >> +#include <linux/regulator/consumer.h>
> >> +
> >> +#define MODULE_NAME "adc084s021"
> >> +#define DRIVER_VERSION "1.0"
> >
> > is only used once at the very end...
> >
> >> +#define ADC_RESOLUTION 8
> Put it inline...
> >> +#define ADC_N_CHANNELS 4
> Belongs inline. No additional info provided by having it here.
> >
> > we want a consistent prefix, such as ADC084S021_
> >
> >> +
> >> +struct adc084s021_configuration {
> >> + const struct iio_chan_spec *channels;
> >> + u8 num_channels;
> >
> > no need for u8, perhaps unsigned?
> >
> >> +};
> >> +
> >> +struct adc084s021 {
> >> + struct spi_device *spi;
> >> + struct spi_message message;
> >> + struct spi_transfer spi_trans[2];
> >> + struct regulator *reg;
> >> + struct mutex lock;
> >> + /*
> >> + * DMA (thus cache coherency maintenance) requires the
> >> + * transfer buffers to live in their own cache lines.
> >> + */
> >> + union {
> >> + u8 tx_buf[2];
> >> + u8 rx_buf[2];
> >> + } ____cacheline_aligned;
> >> + u8 cur_adc_values[ADC_N_CHANNELS];
> >> +};
> >> +
> >> +/**
> >> + * Event triggered when value changes on a channel
> >> + */
> >> +static const struct iio_event_spec adc084s021_event = {
> >> + .type = IIO_EV_TYPE_CHANGE,
> >> + .dir = IIO_EV_DIR_NONE,
> >> +};
> Not the intent of that type of event at all.
> >> +
> >> +/**
> >> + * Channel specification
> >> + */
> >> +#define ADC084S021_VOLTAGE_CHANNEL(num) \
> >> + { \
> >> + .type = IIO_VOLTAGE, \
> >> + .channel = (num), \
> >> + .address = (num << 3), \
> >
> > parenthesis should be around (num)
> >
> >> + .indexed = 1, \
> >> + .scan_index = num, \
> >
> > parenthesis?
> >
> >> + .scan_type = { \
> >> + .sign = 'u', \
> >> + .realbits = 8, \
> >> + .storagebits = 32, \
> >> + .shift = 24 - ((num << 3)), \
> >
> > no need for (( )) around the expression, parenthesis should be around num
> >
> > the shift doesn't make sense, you are shifting in
> >
> > adc084s021_adc_conversion() already and return just 8 bits (so storagebits
> > should be 8)?
> >
> > you could/should use realbits = 8, storagebits = 16, shift = 4 and
> > endianness = IIO_BE if I read Figure 1 of the datasheet correctly
> >
> >> + }, \
> >> + .info_mask_separate = BIT(IIO_CHAN_INFO_RAW), \
> >
> > likely this is missing _SCALE
> > IIO expects data to be returned in millivolt, see
> > Documentation/ABI/testing/sysfs-bus-iio
> >
> >> + .event_spec = &adc084s021_event, \
> >> + .num_event_specs = 1, \
> >
> > ARRAY_SIZE(&adc084s021_event) to be extensible?
> >
> >> + }
> >> +
> >> +static const struct iio_chan_spec adc084s021_channels[] = {
> >> + ADC084S021_VOLTAGE_CHANNEL(0),
> >> + ADC084S021_VOLTAGE_CHANNEL(1),
> >> + ADC084S021_VOLTAGE_CHANNEL(2),
> >> + ADC084S021_VOLTAGE_CHANNEL(3),
> >> + IIO_CHAN_SOFT_TIMESTAMP(4),
> >> +};
> >> +
> >> +static const struct adc084s021_configuration adc084s021_config[] = {
> >> + { adc084s021_channels, ARRAY_SIZE(adc084s021_channels) },
> >
> > for just one configuration, this is not really needed; so you plan/forsee
> > more chips being added soonish?
> >
> >> +};
> >> +
> >> +/**
> >> + * Read an ADC channel and return its value.
> >> + *
> >> + * @adc: The ADC SPI data.
> >> + * @channel: The IIO channel data structure.
> >> + */
> >> +static int adc084s021_adc_conversion(struct adc084s021 *adc,
> >> + struct iio_chan_spec const *channel)
> >> +{
> >> + u16 value;
> >
> > value should be u8, but is not really needed
> >
> >> + int ret;
> >> +
> >> + mutex_lock(&adc->lock);
> >> + adc->tx_buf[0] = channel->address;
> >> +
> >> + /* Do the transfer */
> >> + ret = spi_sync(adc->spi, &adc->message);
> >> +
> >
> > no newline here please
> >
> >> + if (ret < 0) {
> >> + mutex_unlock(&adc->lock);
> >> + return ret;
> >> + }
> >> +
> >> + value = (adc->rx_buf[0] << 4) | (adc->rx_buf[1] >> 4);
> >
> > I recommend using __be16 for rx_buf and
> > ret = (be16_to_cpu(adc->rx_buf) >> 4) & 0xff;
> >
> >> + mutex_unlock(&adc->lock);
> >> +
> >> + dev_dbg(&adc->spi->dev, "value 0x%02X on channel %d\n",
> >> + value, channel->channel);
> >> + return value;
> >> +}
> >> +
> >> +/**
> >> + * Make a readout of requested IIO channel info.
> >
> > no need to document this, it is found in every IIO driver...
> >
> >> + *
> >> + * @indio_dev: The industrial I/O device.
> >> + * @channel: The IIO channel data structure.
> >> + * @val: First element of value (integer).
> >> + * @val2: Second element of value (fractional).
> >> + * @mask: The info_mask to read.
> >> + */
> >> +static int adc084s021_read_raw(struct iio_dev *indio_dev,
> >> + struct iio_chan_spec const *channel, int *val,
> >> + int *val2, long mask)
> >> +{
> >> + struct adc084s021 *adc = iio_priv(indio_dev);
> >> + int retval;
> >
> > how about using ret everywhere?
> Agreed.
> >
> >> +
> >> + switch (mask) {
> >> + case IIO_CHAN_INFO_RAW:
> >
> > probably want iio_device_claim_direct_mode() so to not interfere with
> > buffered accessBorderline case as all you will do is queue up additional spi transfers.
> At least after you have dropped the unusual events stuff.
>
> Arguably this will mean sysfs reads could make the buffered data flow
> less deterministic though so maybe on claiming direct mode (which
> prevents it running concurrently with buffered capture.
> >
> >> + retval = adc084s021_adc_conversion(adc, channel);
> >> + if (retval < 0)
> >> + return retval;
> >> +
> >> + *val = retval;
> >> + return IIO_VAL_INT;
> >> +
> >> + default:
> >> + return -EINVAL;
> >> + }
> >> +}
> >> +
> >> +/**
> >> + * Read enabled ADC channels and push data to the buffer.
> >> + *
> >> + * @irq: The interrupt number (not used).
> >> + * @pollfunc: Pointer to the poll func.
> >> + */
> >> +static irqreturn_t adc084s021_trigger_handler(int irq, void *pollfunc)
> >> +{
> >> + struct iio_poll_func *pf = pollfunc;
> >> + struct iio_dev *indio_dev = pf->indio_dev;
> >> + struct adc084s021 *adc = iio_priv(indio_dev);
> >> + u8 *data;
> >> + s64 timestamp;
> >> + int value, scan_index;
> >> +
> >> + data = kzalloc(indio_dev->scan_bytes, GFP_KERNEL);
> >
> > pre-allocate buffer with maximum size statically; allocation is
> > potentially expensive; don't forget space for the timestamp and padding...
> >
> >> + if (!data) {
> >> + iio_trigger_notify_done(indio_dev->trig);
> >> + return IRQ_NONE;
> >> + }
> >> +
> >> + timestamp = iio_get_time_ns(indio_dev);
> >> +
> >> + for_each_set_bit(scan_index, indio_dev->active_scan_mask,
> >> + indio_dev->masklength) {
> >> + const struct iio_chan_spec *channel =
> >> + &indio_dev->channels[scan_index];
> >> + value = adc084s021_adc_conversion(adc, channel);
> >
> > lock is taken and released for each channel, probably want to do it just
> > once?
> >
> >> + data[scan_index] = value;
> >> +
> >> + /*
> >> + * Compare read data to previous read. If it differs send
> >> + * event notification for affected channel.
> >> + */
> >> + if (adc->cur_adc_values[scan_index] != (u8)value) {
> >
> > cur_adc_values is not initialized (probably set to 0);
> > so on first read, should the notification be sent?
> ? This is 'unusual' to say the least. Why the events given the data
> is available via the buffer. If you really want to do this stuff, it
> ought to be in userspace.
>
> Are you trying to emulate the filtering input does?
The buffer is continuously updated with readings that may have the same
data for many iterations, so my idea was to send an event with timestamp
so that whenever some "new" (changed) data is written there it can be
more easily located in the buffer by the consumer, by matching event
timestamp with buffer timestamp. Is there another way to do this or
should the buffer readings be continuously analysed by another module or
application?
When prototyping this driver I even tweaked the IIO_EVENT_CODE to
include the new value in the event signal itself. But I wasn't brave
enough to show it here :) Please tell me, am I totally wrong by even
trying that idea?
> >
> >> + adc->cur_adc_values[scan_index] = (u8)value;
> >> + iio_push_event(indio_dev,
> >> + IIO_EVENT_CODE(IIO_VOLTAGE, 0,
> >> + IIO_NO_MOD, IIO_EV_DIR_NONE,
> >> + IIO_EV_TYPE_CHANGE,
> >> + channel->channel, 0, 0),
> >> + timestamp);
> >> + dev_dbg(&indio_dev->dev,
> >> + "new value on ch%d: 0x%02X (ts %llu)\n",
> >> + channel->channel, value, timestamp);
> >> + }
> >> + }
> >> +
> >> + iio_push_to_buffers_with_timestamp(indio_dev, data, timestamp);
> >> + iio_trigger_notify_done(indio_dev->trig);
> >> + kfree(data);
> blank line here please.
> >> + return IRQ_HANDLED;
> >> +}
> >> +
> >> +static const struct iio_info adc084s021_info = {
> >> + .read_raw = adc084s021_read_raw,
> >> + .driver_module = THIS_MODULE,
> >> +};
> >> +
> >> +/**
> >> + * Create and register ADC IIO device for SPI.
> >
> > comment is rather pointless
> >
> >> + */
> >> +static int adc084s021_probe(struct spi_device *spi)
> >> +{
> >> + struct iio_dev *indio_dev;
> >> + struct adc084s021 *adc;
> >> + int config = spi_get_device_id(spi)->driver_data;
> >> + int retval;
> >
> > ret maybe
> >
> >> +
> >> + /* Allocate an Industrial I/O device */
> >
> > obviously...
> :) Yeah, clear out any comments that don't tell us much.
> >
> >> + indio_dev = devm_iio_device_alloc(&spi->dev, sizeof(*adc));
> >> + if (!indio_dev) {
> >> + dev_err(&spi->dev, "Failed to allocate IIO device\n");
> >> + return -ENOMEM;
> >> + }
> >> +
> >> + adc = iio_priv(indio_dev);
> >> + adc->spi = spi;
> >> + spi->bits_per_word = ADC_RESOLUTION;
> This surprised me somewhat as I was expecting an odd value for this
> (and was going to ask if standard power of 2 options would work).
> If it had been inline, this would have required slightly less review
> time which I always like (particularly when crammed into an airline seat!)
> >> +
> >> + /* Update the SPI device with config and connect the iio dev */
> >> + retval = spi_setup(spi);
> >> + if (retval) {
> >> + dev_err(&spi->dev, "Failed to update SPI device\n");
> >> + return retval;
> >> + }
> >> + spi_set_drvdata(spi, indio_dev);
> >> +
> >> + /* Initiate the Industrial I/O device */
> :>> + indio_dev->dev.parent = &spi->dev;
> >> + indio_dev->dev.of_node = spi->dev.of_node;
> >> + indio_dev->name = spi_get_device_id(spi)->name;
> >> + indio_dev->modes = INDIO_DIRECT_MODE;
> >> + indio_dev->info = &adc084s021_info;
> >> + indio_dev->channels = adc084s021_config[config].channels;
> >> + indio_dev->num_channels = adc084s021_config[config].num_channels;
> >
> > could do it directly as long there is just one configuration
> That would certainly be preferable unless V2 is going to show up
> with multiple options!
> >
> >> +
> >> + /* Create SPI transfer for channel reads */
> >> + adc->spi_trans[0].tx_buf = &adc->tx_buf[0];
> >> + adc->spi_trans[0].len = 2;
> >> + adc->spi_trans[0].speed_hz = spi->max_speed_hz;
> >> + adc->spi_trans[0].bits_per_word = spi->bits_per_word;
> >> + adc->spi_trans[1].rx_buf = &adc->rx_buf[0];
> >> + adc->spi_trans[1].len = 2;
> >> + adc->spi_trans[1].speed_hz = spi->max_speed_hz;
> >> + adc->spi_trans[1].bits_per_word = spi->bits_per_word;
> >> +
> >> + /* Setup SPI message for channel reads */
> >> + spi_message_init(&adc->message);
> >> + spi_message_add_tail(&adc->spi_trans[0], &adc->message);
> >> + spi_message_add_tail(&adc->spi_trans[1], &adc->message);
> spi_init_with_transfers to save a bit of boilerplate.
> >> +
> >> + adc->reg = devm_regulator_get(&spi->dev, "vref");
> >> + if (IS_ERR(adc->reg))
> >> + return PTR_ERR(adc->reg);
> >> +
> >> + retval = regulator_enable(adc->reg);
> >> + if (retval < 0)
> >> + return retval;
> Given we either have slow sysfs accesses or know we have buffered
> access on going. Have you considered enabling and disabling
> the regulator dynamically? The fact that this often makes sense is
> why Mark and co from the regulators side of things haven't provided
> a devm_regulator_enable... You would need a preenable and postdisable
> to make sure it was on for buffered access.
> >> +
> >> + mutex_init(&adc->lock);
> >> +
> >> + /* Setup triggered buffer with pollfunction */
> >> + retval = iio_triggered_buffer_setup(indio_dev, NULL,
> >
> > devm_()
> I disagree. It would change the ordering wrt to the regulator_enable.
> Now, obviously it won't actually matter, but it will make the code
> less obviously correct and for the small burden of extra unwind
> code I'd keep using the non devm version.
> >
> >> + adc084s021_trigger_handler, NULL);
> >> + if (retval) {
> >> + dev_err(&spi->dev, "Failed to setup triggered buffer\n");
> >> + goto buffer_setup_failed;
> >> + }
> >> +
> >> + retval = iio_device_register(indio_dev);
> >> + if (retval) {
> >> + dev_err(&spi->dev, "Failed to register IIO device\n");
> Hmm. If we were to flesh out some error messages for the few
> cases where there is no info provided in iio_device_register we could
> probably clean out a fair bit of boiler plate reporting in drivers.
> Ah well, one for the future!
> >> + goto device_register_failed;
> >> + }
> >> +
> >> + dev_info(&spi->dev, "probed!\n");
> >
> > no logging please, just outputs clutter
> >
> >> + return 0;
> >> +
> >> +device_register_failed:
> >> + iio_triggered_buffer_cleanup(indio_dev);
> >> +buffer_setup_failed:
> >> + regulator_disable(adc->reg);
> >> + return retval;
> >> +}
> >> +
> >> +/**
> >> + * Unregister ADC IIO device for SPI.
> >> + */
> >> +static int adc084s021_remove(struct spi_device *spi)
> >> +{
> >> + struct iio_dev *indio_dev = spi_get_drvdata(spi);
> >> + struct adc084s021 *adc = iio_priv(indio_dev);
> >> +
> >> + iio_device_unregister(indio_dev);
> >> + iio_triggered_buffer_cleanup(indio_dev);
> >> + regulator_disable(adc->reg);
> blank line here preferred (slightly!)
> >> + return 0;
> >> +}
> >> +
> >> +static const struct of_device_id adc084s021_of_match[] = {
> >> + { .compatible = "ti,adc084s021", },
> >> + {},
> >> +};
> >> +
> >> +MODULE_DEVICE_TABLE(of, adc084s021_of_match);
> >> +
> >> +static const struct spi_device_id adc084s021_id[] = {
> >> + { MODULE_NAME, 0},
> >> + {}
> >> +};
> >> +
> >> +MODULE_DEVICE_TABLE(spi, adc084s021_id);
> >> +
> >> +static struct spi_driver adc084s021_driver = {
> >> + .driver = {
> >> + .name = MODULE_NAME,
> >> + .of_match_table = of_match_ptr(adc084s021_of_match),
> >> + },
> >> + .probe = adc084s021_probe,
> >> + .remove = adc084s021_remove,
> >> + .id_table = adc084s021_id,
> >> +};
> >> +
> Convention often says to not bother with a blank line here or before
> the MODULE_DEVICE_TABLE above. Gives a visual indication that these macros
> are being passed the structures.
> (really minor point!)
> >> +module_spi_driver(adc084s021_driver);
> >> +
> >> +MODULE_AUTHOR("Mårten Lindahl <martenli@xxxxxxxx>");
> >> +MODULE_DESCRIPTION("Texas Instruments ADC084S021");
> >> +MODULE_LICENSE("GPL v2");
> >> +MODULE_VERSION(DRIVER_VERSION);
> >>
> >
>
--
To unsubscribe from this list: send the line "unsubscribe devicetree" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
