Re: [PATCH v3 1/7] iio: adc: hi8435: Holt HI-8435 threshold detector

From: Jonathan Cameron
Date: Sun Aug 16 2015 - 05:01:11 EST


On 11/08/15 15:37, Vladimir Barinov wrote:
> Hi Jonathan,
>
> Thank you for the review.
Sorry I was being so indecisive. Always take a while when something
'new' turns up to gather opinions and select the best option. You
are just the unlucky guy who happened to hit the question first!

Anyhow, I'm basically happy (now) with the events approach though
I would suggest some modifications to how exactly it is done
(see inline).

Jonathan
>
> On 08.08.2015 20:56, Jonathan Cameron wrote:
>> On 29/07/15 13:56, Vladimir Barinov wrote:
>>> Add Holt threshold detector driver for HI-8435 chip
>>>
>>> Signed-off-by: Vladimir Barinov <vladimir.barinov@xxxxxxxxxxxxxxxxxx>
>> Because it usually reads better that way, I review from the bottom.
>> Might make more sense if you read the comments that way around ;)
> I did try to match this way.
> At some places the comments below the code, but in other places the comments are above the code :)
Yeah, I've never been consistent ;)
>>
>> I'm going to guess that the typical usecase for this device is in realtime
>> systems where polling gives a nice predictable timing (hence the lack of any
>> interrupt support). These typically operate as 'scanning' devices
>> (e.g. you update all state at approximately the same time, by reading
>> one input after another in some predefined order).
> Probably you are right about 8435 h/w design or there were other
> reason why this functionality was dropped in comparison to 8429
> chip. The Holt chip hi-8429 has both hardware interrupt and hardware
> debounce enable lines, but it has less amount of sensors (just 8
> outputs).
>>
>> Does the use of events actually map well to this use case? You are taking
>> something that (other than the results being a bit different) is basically
>> a digital I/O interface and mapping it (sort of) to an interrupt chip
>> when it isn't nothing of the sort.
> a) Using hardware point of view.
>
> I was thinking that it matches the events a little bit more then
> buffer, because the chip senses for threshold changes (passing
> threshold high or threshold low margins) and then notifies the upper
> layer. The bad point here is that the chip does not have interrupt
> line for sensor state change (i.e. threshold passing).
>
> If it would have the interrupt line for sensor state changes then the
> event interface map this case best way.
Agreed. I'm coming round to the argument Lars made that we should end
up with the interface being the same whether or not there is an interrupt
there. Actually I suspect in the long run we will end up with both
a buffered interface and an event interface as they simply have different
usecases for the same hardware (which is why we are having these lengthy
discussions in the first place!)

> b) Using s/w point of view
>
> I answer relying on current iio stack implementation. I do understand
> that the 8435 chip is not the common/usual iio client like
> adc/dac/light/others, so I'm trying to match the existent iio stack.
Yes. It's certainly throwing up lots of questions!
>
> From s/w point of view it does not matter much difference between
> event or buffer interface for this chip because in provides 1-bit
> change and does not have it's own interrupt source line.
Yes, just requires the userspace side to track the state if it wants
to know. Actually the one nasty is initialization. When the events
are first enabled I guess we fire them all once (if any limits are currently
breached). That way the state can always be known.
>
> The event interface has generic interface to fill in falling/rising
> thresholds but buffer interface does not.
We can always add more ABI.

> The buffer interface has already has the trigger poll func support
> but event interface does not. So I've tried to implement the
> triggered event in iio stack. BTW: Doesn't the triggered event
> support match the usecase with iio_interrupt_trigger? Or it does not
> make sense to have triggered events (even with irqtrig) at all?
Initially I was unsure about this, but Lars and you have convinced me.
I'm happy having that in the subsystem. Can also be used for devices
which have an interrupt but which didn't get wired for some reason rather
than having them have some internal poll loop (as we have done up to now).


>>
>> I'd like more opinions on this, but my gut reaction is that we would
>> be better making the normal buffered infrastructure handle this data
>> properly rather than pushing it through the events intrastructure.
>>
>> Your solution is neat and tidy in someways but feels like it is mashing
>> data into a particular format.
> Probably that's all because the chip lacks smth from buffer interface
> approach (the data should diferent the 1-bit) and smth from event
> interface approach (lack of h/w interrupt line)
Agreed.
>>
>> To add to the complexity we could have debounce logic. If we mapped to a
>> fill the buffer with a scan mode, how would we handle this?
>> My guess is that it would simply delay transistions. Perhaps we even
>> support reading both the value right now and the debounced value if that
>> is possible?

> I did handle it in the driver the same way in both cases (buffer and
> event interfaces): event/buffer trigger issues interrupt and then the
> driver checks value right now and after delay (debounce_interval).
I am yet to be convinced we want software debounce in the kernel. This
is something that the input subsystem has always had. Partly the aim of
IIO was to provide a slim system where all this stuff was moved to userspace.

Ultimately the ideal option to my mind is to finally write an inkernel
interface for the events data flow (as we have for buffered data) then
we can have a generic attachable debounce unit independent of any particular
driver. Hohum. Now all I need is someone to write it or a few weeks off
work ;)
> If we do not plan to put debounced values to buffer then we may cache
> during debouncing procedure and then return via
> IIO_CHAN_INFO_DEBOUNCE_VALUE (or other info name). Probably even with
> timestamp.

>>
>> However, here the debounce is all software. To my mind, move this into
>> userspace entirely. We aren't dealing with big data here so it's hardly
>> going to be particularly challenging.
> I will drop it on your demand in favour to encourage chipmakers to provide h/w debounce,
> but this was much easy/efficient to have it in kernel space, since
> during run generic_buffer sample (or other) we got the new data and then we have to start timer, then timer expires
> and we read raw values from sysfs manually to compare the one that came from buffer and one after debounce
> delay.
If you are clocking evenly and we were using buffering you'd simply do it as readback in time
rather than the other way around - thus no timer at all. Just each time check the last N
samples to see if they are all high or all low or similar.

>
>>
>> So my gut feeling is definitely we need to make the buffer infrastructure
>> handle 1 bit values (in groups) then push this data out that way.
> I will wait for your final decision.
Lets proceed with the event version. If someone has a strong usecase
for adding buffering as well, there is nothing stopping the driver doing both
that I can see (once we have the core infrastructure in place).
>
>>
>> Still I would like other opinions on this!
>>
>> Jonathan
>>
>>> ---
>>> Changes in version 2:
>>> - Added file sysfs-bus-iio-adc-hi8435
>>> - Changed naming from "discrete ADC" to "threshold detector"
>>> - Replaced swab16p/swab32p with be16_to_cpup/be32_to_cpup
>>> - Made *_show and *_store functions static
>>> - moved out from iio buffers to iio events
>>> - removed hi8436/hi8437 chips from the driver
>>> - moved from debounce_soft_delay/enable to debounce_interval via
>>> IIO_CHAN_INFO_DEBOUNCE_TIME
>>> - added name extention "comparator"
>>> - moved threshold/hysteresis setup via generic iio event sysfs
>>> - added software mask/unmask channel events
>>> - added programming sensor outputs while in test mode via
>>> IIO_CHAN_INFO_RAW
>>> - added channels .ext_info for programming sensing mode
>>> Changes in version 3:
>>> - fixed typos
>>> - moved <linux/interrupt.h> to match alphabetic order
>>> - fixed missed */event/* prefix in the ABI description
>>> - added missed colon in sysfs-bus-iio-adc-hi8435 file after "What"
>>> - moved in_voltageY_comparator_thresh_either_en and debounce_time
>>> to sysfs-bus-iio file
>>> - added error checking for hi8435_write[b|w]
>>>
>>> Documentation/ABI/testing/sysfs-bus-iio | 1 +
>>> Documentation/ABI/testing/sysfs-bus-iio-adc-hi8435 | 61 ++
>>> drivers/iio/adc/Kconfig | 11 +
>>> drivers/iio/adc/Makefile | 1 +
>>> drivers/iio/adc/hi8435.c | 659 +++++++++++++++++++++
>>> 5 files changed, 742 insertions(+)
>>> create mode 100644 Documentation/ABI/testing/sysfs-bus-iio-adc-hi8435
>>> create mode 100644 drivers/iio/adc/hi8435.c
>>>
>>> diff --git a/Documentation/ABI/testing/sysfs-bus-iio b/Documentation/ABI/testing/sysfs-bus-iio
>>> index 70c9b1a..09eac44 100644
>>> --- a/Documentation/ABI/testing/sysfs-bus-iio
>>> +++ b/Documentation/ABI/testing/sysfs-bus-iio
>>> @@ -572,6 +572,7 @@ What: /sys/.../iio:deviceX/events/in_rot_from_north_magnetic_tilt_comp_thresh_r
>>> What: /sys/.../iio:deviceX/events/in_rot_from_north_magnetic_tilt_comp_thresh_falling_en
>>> What: /sys/.../iio:deviceX/events/in_rot_from_north_true_tilt_comp_thresh_rising_en
>>> What: /sys/.../iio:deviceX/events/in_rot_from_north_true_tilt_comp_thresh_falling_en
>>> +What: /sys/.../iio:deviceX/events/in_voltageY_comparator_thresh_either_en
>>> What: /sys/.../iio:deviceX/events/in_voltageY_supply_thresh_rising_en
>>> What: /sys/.../iio:deviceX/events/in_voltageY_supply_thresh_falling_en
>>> What: /sys/.../iio:deviceX/events/in_voltageY_thresh_rising_en
>>> diff --git a/Documentation/ABI/testing/sysfs-bus-iio-adc-hi8435 b/Documentation/ABI/testing/sysfs-bus-iio-adc-hi8435
>>> new file mode 100644
>>> index 0000000..c59d81d
>>> --- /dev/null
>>> +++ b/Documentation/ABI/testing/sysfs-bus-iio-adc-hi8435
>>> @@ -0,0 +1,61 @@
>>> +What: /sys/bus/iio/devices/iio:deviceX/in_voltageY_comparator_raw
We really try to use the extended names infrequently and normally only
for things like 'this voltage channel is internally connected to the power
supply rail'. Don't think we need it here. As suggested below, if we
are going events then we don't want to expose a _raw reading except via
in the 'events' directory where it would be a 'current event status'.
Anyhow, I talk about that futher down.

>>> +Date: July 2015
>>> +KernelVersion: 4.2.0
>>> +Contact: source@xxxxxxxxxxxxxxxxxx
>>> +Description:
>>> + Read value is a voltage threshold measurement from channel Y.
>>> + Could be either 0 if sensor voltage is lower then low voltage
>>> + threshold or 1 if sensor voltage is higher then high voltage
>>> + threshold.
>>> + Write value is a programmed sensor output while in self test
>>> + mode. Could be either 0 or 1. The programmed value will be read
>>> + back if /sys/bus/iio/devices/iio:deviceX/test_enable is set to 1.
>>> +
>>> +What: /sys/bus/iio/devices/iio:deviceX/test_enable
>>> +Date: July 2015
>>> +KernelVersion: 4.2.0
>>> +Contact: source@xxxxxxxxxxxxxxxxxx
>>> +Description:
>>> + Enable/disable the HI-8435 self test mode.
>>> + If enabled the in_voltageY_comparator_raw should be read back
>>> + accordingly to written value to in_voltageY_comparator_raw.

Hmm. Normally we just do self tests at startup. So set some patterns, read them
and then verify they are fine. Drops the need for a userspace interface
that can get a little confusing (writing to what is normally a read
only attribute). Would this work here?

Alternatively, just add a debugfs interface to allow this to be set rather than
exposing it via sysfs all the time?

>>> +
>>> +What: /sys/bus/iio/devices/iio:deviceX/in_voltageY_comparator_sensing_mode
>>> +Date: July 2015
>>> +KernelVersion: 4.2.0
>>> +Contact: source@xxxxxxxxxxxxxxxxxx
>>> +Description:
>>> + Program sensor type for threshold detector inputs.
>>> + Could be either "GND-Open" or "Supply-Open" mode. Y is a
>>> + threshold detector input channel. Channels 0..7, 8..15, 16..23
>>> + and 24..31 has common sensor types.
>>> +
>>> +What: /sys/bus/iio/devices/iio:deviceX/events/in_voltageY_comparator_thresh_falling_value
This is where the comparator extended name looks a bit silly as we are
clearly comparing it anyway due to the thresh_falling part ;)
>>> +Date: July 2015
>>> +KernelVersion: 4.2.0
>>> +Contact: source@xxxxxxxxxxxxxxxxxx
>>> +Description:
>>> + Channel Y low voltage threshold. If sensor input voltage goes lower then
>>> + this value then the threshold falling event is pushed.
>>> + Depending on in_voltageY_comparator_sensing_mode the low voltage threshold
>>> + is separately set for "GND-Open" and "Supply-Open" modes.
>>> + Channels 0..31 have common low threshold values, but could have different
>>> + sensing_modes.
>>> + The low voltage threshold range is between 2..21V.
>>> + Hysteresis between low and high thresholds can not be lower then 2 and
>>> + can not be odd.
>>> +
>>> +What: /sys/bus/iio/devices/iio:deviceX/events/in_voltageY_comparator_thresh_rising_value
>>> +Date: July 2015
>>> +KernelVersion: 4.2.0
>>> +Contact: source@xxxxxxxxxxxxxxxxxx
>>> +Description:
>>> + Channel Y high voltage threshold. If sensor input voltage goes higher then
>>> + this value then the threshold rising event is pushed.
>>> + Depending on in_voltageY_comparator_sensing_mode the high voltage threshold
>>> + is separately set for "GND-Open" and "Supply-Open" modes.
>>> + Channels 0..31 have common high threshold values, but could have different
>>> + sensing_modes.
>>> + The high voltage threshold range is between 3..22V.
>>> + Hysteresis between low and high thresholds can not be lower then 2 and
>>> + can not be odd.
>>> diff --git a/drivers/iio/adc/Kconfig b/drivers/iio/adc/Kconfig
>>> index eb0cd89..553c91e 100644
>>> --- a/drivers/iio/adc/Kconfig
>>> +++ b/drivers/iio/adc/Kconfig
>>> @@ -170,6 +170,17 @@ config EXYNOS_ADC
>>> of SoCs for drivers such as the touchscreen and hwmon to use to share
>>> this resource.
>>> +config HI8435
>>> + tristate "Holt Integrated Circuits HI-8435 threshold detector"
>>> + select IIO_TRIGGERED_EVENT
>>> + depends on SPI
>>> + help
>>> + If you say yes here you get support for Holt Integrated Circuits
>>> + HI-8435 chip.
>>> +
>>> + This driver can also be built as a module. If so, the module will be
>>> + called hi8435.
>>> +
>>> config LP8788_ADC
>>> tristate "LP8788 ADC driver"
>>> depends on MFD_LP8788
>>> diff --git a/drivers/iio/adc/Makefile b/drivers/iio/adc/Makefile
>>> index a096210..00f367aa 100644
>>> --- a/drivers/iio/adc/Makefile
>>> +++ b/drivers/iio/adc/Makefile
>>> @@ -19,6 +19,7 @@ obj-$(CONFIG_BERLIN2_ADC) += berlin2-adc.o
>>> obj-$(CONFIG_DA9150_GPADC) += da9150-gpadc.o
>>> obj-$(CONFIG_CC10001_ADC) += cc10001_adc.o
>>> obj-$(CONFIG_EXYNOS_ADC) += exynos_adc.o
>>> +obj-$(CONFIG_HI8435) += hi8435.o
>>> obj-$(CONFIG_LP8788_ADC) += lp8788_adc.o
>>> obj-$(CONFIG_MAX1027) += max1027.o
>>> obj-$(CONFIG_MAX1363) += max1363.o
>>> diff --git a/drivers/iio/adc/hi8435.c b/drivers/iio/adc/hi8435.c
>>> new file mode 100644
>>> index 0000000..f3dab5a
>>> --- /dev/null
>>> +++ b/drivers/iio/adc/hi8435.c
>>> @@ -0,0 +1,659 @@
>>> +/*
>>> + * Holt Integrated Circuits HI-8435 threshold detector driver
>>> + *
>>> + * Copyright (C) 2015 Zodiac Inflight Innovations
>>> + * Copyright (C) 2015 Cogent Embedded, Inc.
>>> + *
>>> + * This program is free software; you can redistribute it and/or modify it
>>> + * under the terms of the GNU General Public License as published by the
>>> + * Free Software Foundation; either version 2 of the License, or (at your
>>> + * option) any later version.
>>> + */
>>> +
>>> +#include <linux/delay.h>
>>> +#include <linux/iio/events.h>
>>> +#include <linux/iio/iio.h>
>>> +#include <linux/iio/sysfs.h>
>>> +#include <linux/iio/trigger.h>
>>> +#include <linux/iio/trigger_consumer.h>
>>> +#include <linux/iio/triggered_event.h>
>>> +#include <linux/interrupt.h>
>>> +#include <linux/module.h>
>>> +#include <linux/of.h>
>>> +#include <linux/of_device.h>
>>> +#include <linux/of_gpio.h>
>>> +#include <linux/spi/spi.h>
>>> +#include <linux/workqueue.h>
>>> +
>>> +#define DRV_NAME "hi8435"
>>> +
>>> +/* Register offsets for HI-8435 */
>>> +#define HI8435_CTRL_REG 0x02
>>> +#define HI8435_PSEN_REG 0x04
>>> +#define HI8435_TMDATA_REG 0x1E
>>> +#define HI8435_GOCENHYS_REG 0x3A
>>> +#define HI8435_SOCENHYS_REG 0x3C
>> Hmm. These aren't in the docs that I can immediately acquire...
>> Any chance of some more meaningful naming?
> I did follow the register naming in the chip datasheet:
> http://www.holtic.com/products/3081-hi-8435.aspx
>
> The SO7_0 means "Sensor status bank 0 register: SO<7:0>".
> Does HI8435_STATUS_BANK0_REG work?
Ah, I missed the meaning entirely by not noticing they were 8 bits
each ;) This current naming is fine.
> Or should I just add comments instead of changing definition?
>>> +#define HI8435_SO7_0_REG 0x10
>>> +#define HI8435_SO15_8_REG 0x12
>> +#define HI8435_SO23_16_REG 0x14
>>> +#define HI8435_SO31_24_REG 0x16
>>> +#define HI8435_SO31_0_REG 0x78
>>> +
>>> +#define HI8435_WRITE_OPCODE 0x00
>>> +#define HI8435_READ_OPCODE 0x80
>>> +
>>> +/* CTRL register bits */
>>> +#define HI8435_CTRL_TEST 0x01
>>> +#define HI8435_CTRL_SRST 0x02
>>> +
>>> +#define HI8435_DEBOUNCE_DELAY_MAX 1000 /* msec */
>>> +#define HI8435_DEBOUNCE_DELAY_DEF 100 /* msec */
>>> +
>>> +struct hi8435_priv {
>>> + struct spi_device *spi;
>>> + struct mutex lock;
>>> + struct delayed_work work;
>>> +
>>> + int reset_gpio;
>>> + int debounce_interval; /* msec */
>>> + u32 debounce_val; /* prev value to compare during software debounce */
>>> +
>>> + unsigned long event_scan_mask; /* soft mask/unmask channels events */
>>> + unsigned int event_prev_val;
>>> +
>>> + unsigned threshold_lo[2]; /* GND-Open and Supply-Open thresholds */
>>> + unsigned threshold_hi[2]; /* GND-Open and Supply-Open threshold */
>>> + u8 reg_buffer[4] ____cacheline_aligned;
>>> +};
>>> +
>>> +static int hi8435_readb(struct hi8435_priv *priv, u8 reg, u8 *val)
>>> +{
>>> + reg |= HI8435_READ_OPCODE;
>>> + return spi_write_then_read(priv->spi, &reg, 1, val, 1);
>>> +}
>>> +
>>> +static int hi8435_readw(struct hi8435_priv *priv, u8 reg, u16 *val)
>>> +{
>>> + int ret;
>>> +
>>> + reg |= HI8435_READ_OPCODE;
>>> + ret = spi_write_then_read(priv->spi, &reg, 1, val, 2);
>>> + *val = be16_to_cpup(val);
>>> +
>>> + return ret;
>>> +}
>>> +
>>> +static int hi8435_readl(struct hi8435_priv *priv, u8 reg, u32 *val)
>>> +{
>>> + int ret;
>>> +
>>> + reg |= HI8435_READ_OPCODE;
>>> + ret = spi_write_then_read(priv->spi, &reg, 1, val, 4);
>>> + *val = be32_to_cpup(val);
>>> +
>>> + return ret;
>>> +}
>>> +
>>> +static int hi8435_writeb(struct hi8435_priv *priv, u8 reg, u8 val)
>>> +{
>>> + priv->reg_buffer[0] = reg | HI8435_WRITE_OPCODE;
>>> + priv->reg_buffer[1] = val;
>>> +
>>> + return spi_write(priv->spi, priv->reg_buffer, 2);
>>> +}
>>> +
>>> +static int hi8435_writew(struct hi8435_priv *priv, u8 reg, u16 val)
>>> +{
>>> + priv->reg_buffer[0] = reg | HI8435_WRITE_OPCODE;
>>> + priv->reg_buffer[1] = (val >> 8) & 0xff;
>>> + priv->reg_buffer[2] = val & 0xff;
>>> +
>>> + return spi_write(priv->spi, priv->reg_buffer, 3);
>>> +}
>>> +
>>> +static ssize_t hi8435_test_enable_show(struct device *dev,
>>> + struct device_attribute *attr, char *buf)
>>> +{
>>> + struct hi8435_priv *priv = iio_priv(dev_to_iio_dev(dev));
>>> + int ret;
>>> + u8 reg;
>>> +
>>> + ret = hi8435_readb(priv, HI8435_CTRL_REG, &reg);
>>> + if (ret < 0)
>>> + return ret;
>>> +
>>> + return sprintf(buf, "%d\n", reg & HI8435_CTRL_TEST);
>>> +}
>>> +
>>> +static ssize_t hi8435_test_enable_store(struct device *dev,
>>> + struct device_attribute *attr,
>>> + const char *buf, size_t len)
>>> +{
>>> + struct hi8435_priv *priv = iio_priv(dev_to_iio_dev(dev));
>>> + unsigned int val;
>>> + int ret;
>>> +
>>> + ret = kstrtouint(buf, 10, &val);
>>> + if (ret)
>>> + return ret;
>>> +
>>> + ret = hi8435_writeb(priv, HI8435_CTRL_REG, val ? HI8435_CTRL_TEST : 0);
>>> + if (ret < 0)
>>> + return ret;
>>> +
>>> + return len;
>>> +}
>>> +
>>> +static IIO_DEVICE_ATTR(test_enable, S_IRUGO | S_IWUSR,
>>> + hi8435_test_enable_show, hi8435_test_enable_store, 0);
>>> +
>>> +static struct attribute *hi8435_attributes[] = {
>>> + &iio_dev_attr_test_enable.dev_attr.attr,
>>> + NULL,
>>> +};
>>> +
>>> +static struct attribute_group hi8435_attribute_group = {
>>> + .attrs = hi8435_attributes,
>>> +};
>>> +
>>> +static int hi8435_read_raw(struct iio_dev *idev,
>>> + const struct iio_chan_spec *chan,
>>> + int *val, int *val2, long mask)
>>> +{
>>> + struct hi8435_priv *priv = iio_priv(idev);
>>> + int ret;
>>> +
>>> + switch (mask) {
>>> + case IIO_CHAN_INFO_RAW:
>>> + ret = hi8435_readl(priv, HI8435_SO31_0_REG, val);
>>> + if (ret < 0)
>>> + return ret;
>>> + *val = !!(*val & BIT(chan->channel));
>>> + return IIO_VAL_INT;

This is where we end up in an odd possition having elected to report these
as 'events'.
We are reporting a 'last event state' via a read raw rather than as part
of the events infrastructure. I wonder if instead we want to allow the
events stuff to give us a 'am I true now signal'.
Would be easy enough to add and would give a clear interface.

IIO_EV_INFO_CURRENT_STATE and attibute naming something like
in_voltageY_thresh_rising_state
which reads either 1 or 0?

For other devices we might need to take into account the
we are not currently measuring but that can be done by an error
return.

>>> + case IIO_CHAN_INFO_DEBOUNCE_TIME:
>>> + *val = priv->debounce_interval;
>>> + return IIO_VAL_INT;
>>> + default:
>>> + return -EINVAL;
>>> + }
>>> +}
>>> +
>>> +static int hi8435_write_raw(struct iio_dev *idev,
>>> + const struct iio_chan_spec *chan,
>>> + int val, int val2, long mask)
>>> +{
>>> + struct hi8435_priv *priv = iio_priv(idev);
>>> +
>>> + switch (mask) {
>>> + case IIO_CHAN_INFO_RAW:
>>> + /* program sensors outputs in test mode */
>>> + return hi8435_writeb(priv, HI8435_TMDATA_REG, val ? 0x1 : 0x2);
>>> + case IIO_CHAN_INFO_DEBOUNCE_TIME:
>>> + if (val < 0)
>>> + return -EINVAL;
>>> + if (val > HI8435_DEBOUNCE_DELAY_MAX)
>>> + val = HI8435_DEBOUNCE_DELAY_MAX;
>>> + priv->debounce_interval = val;
>>> + return 0;
>>> + default:
>>> + return -EINVAL;
>>> + }
>>> +}
>>> +
>>> +static int hi8435_read_event_config(struct iio_dev *idev,
>>> + const struct iio_chan_spec *chan,
>>> + enum iio_event_type type,
>>> + enum iio_event_direction dir)
>>> +{
>>> + struct hi8435_priv *priv = iio_priv(idev);
>>> +
>>> + return !!(priv->event_scan_mask & BIT(chan->channel));
>>> +}
>>> +
>>> +static int hi8435_write_event_config(struct iio_dev *idev,
>>> + const struct iio_chan_spec *chan,
>>> + enum iio_event_type type,
>>> + enum iio_event_direction dir, int state)
>>> +{
>>> + struct hi8435_priv *priv = iio_priv(idev);
>>> +
>>> + priv->event_scan_mask &= ~BIT(chan->channel);
>>> + if (state)
>>> + priv->event_scan_mask |= BIT(chan->channel);
>>> +
>>> + return 0;
>>> +}
>>> +
>>> +static int hi8435_read_event_value(struct iio_dev *idev,
>>> + const struct iio_chan_spec *chan,
>>> + enum iio_event_type type,
>>> + enum iio_event_direction dir,
>>> + enum iio_event_info info,
>>> + int *val, int *val2)
>>> +{
>>> + struct hi8435_priv *priv = iio_priv(idev);
>>> + int ret;
>>> + u8 mode, psen;
>>> + u16 reg;
>>> +
>>> + ret = hi8435_readb(priv, HI8435_PSEN_REG, &psen);
>>> + if (ret < 0)
>>> + return ret;
>>> +
>>> + /* Supply-Open or GND-Open sensing mode */
>>> + mode = !!(psen & BIT(chan->channel / 8));
>>> +
>>> + ret = hi8435_readw(priv, mode ? HI8435_SOCENHYS_REG :
>>> + HI8435_GOCENHYS_REG, &reg);
>>> + if (ret < 0)
>>> + return ret;
>>> +
>>> + if (dir == IIO_EV_DIR_FALLING)
>>> + *val = ((reg & 0xff) - (reg >> 8)) / 2;
>>> +
>>> + if (dir == IIO_EV_DIR_RISING)
>>> + *val = ((reg & 0xff) + (reg >> 8)) / 2;
>>> +
>>> + return IIO_VAL_INT;
>>> +}
>>> +
>>> +static int hi8435_write_event_value(struct iio_dev *idev,
>>> + const struct iio_chan_spec *chan,
>>> + enum iio_event_type type,
>>> + enum iio_event_direction dir,
>>> + enum iio_event_info info,
>>> + int val, int val2)
>>> +{
>>> + struct hi8435_priv *priv = iio_priv(idev);
>>> + int ret;
>>> + u8 mode, psen;
>>> + u16 reg;
>>> +
>>> + ret = hi8435_readb(priv, HI8435_PSEN_REG, &psen);
>>> + if (ret < 0)
>>> + return ret;
>>> +
>>> + /* Supply-Open or GND-Open sensing mode */
>>> + mode = !!(psen & BIT(chan->channel / 8));
>>> +
>>> + ret = hi8435_readw(priv, mode ? HI8435_SOCENHYS_REG :
>>> + HI8435_GOCENHYS_REG, &reg);
>>> + if (ret < 0)
>>> + return ret;
>>> +
>>> + if (dir == IIO_EV_DIR_FALLING) {
>>> + /* falling threshold range 2..21V, hysteresis minimum 2V */
>>> + if (val < 2 || val > 21 || (val + 1) >= priv->threshold_hi[mode])
>>> + return -EINVAL;
>>> +
>>> + if (val == priv->threshold_lo[mode])
>>> + return 0;
>>> +
>>> + priv->threshold_lo[mode] = val;
>>> +
>>> + /* hysteresis must not be odd */
>>> + if ((priv->threshold_hi[mode] - priv->threshold_lo[mode]) % 2)
>>> + priv->threshold_hi[mode]--;
>>> + }
>>> +
>>> + if (dir == IIO_EV_DIR_RISING) {
>>> + /* rising threshold range 3..22V, hysteresis minimum 2V */
>>> + if (val < 3 || val > 22 || val <= (priv->threshold_lo[mode] + 1))
>>> + return -EINVAL;
>>> +
>>> + if (val == priv->threshold_hi[mode])
>>> + return 0;
>>> +
>>> + priv->threshold_hi[mode] = val;
>>> +
>>> + /* hysteresis must not be odd */
>>> + if ((priv->threshold_hi[mode] - priv->threshold_lo[mode]) % 2)
>>> + priv->threshold_lo[mode]++;
>>> + }
>>> +
>>> + /* program thresholds */
>>> + mutex_lock(&priv->lock);
>>> +
>>> + ret = hi8435_readw(priv, mode ? HI8435_SOCENHYS_REG :
>>> + HI8435_GOCENHYS_REG, &reg);
>>> + if (ret < 0) {
>>> + mutex_unlock(&priv->lock);
>>> + return ret;
>>> + }
>>> +
>>> + /* hysteresis */
>>> + reg = priv->threshold_hi[mode] - priv->threshold_lo[mode];
>>> + reg <<= 8;
>>> + /* threshold center */
>>> + reg |= (priv->threshold_hi[mode] + priv->threshold_lo[mode]);
>>> +
>>> + ret = hi8435_writew(priv, mode ? HI8435_SOCENHYS_REG :
>>> + HI8435_GOCENHYS_REG, reg);
>>> +
>>> + mutex_unlock(&priv->lock);
>>> +
>>> + return ret;
>>> +}
>>> +
>>> +static const struct iio_event_spec hi8435_events[] = {
>>> + {
>>> + .type = IIO_EV_TYPE_THRESH,
>>> + .dir = IIO_EV_DIR_RISING,
>>> + .mask_separate = BIT(IIO_EV_INFO_VALUE),
>>> + }, {
>>> + .type = IIO_EV_TYPE_THRESH,
>>> + .dir = IIO_EV_DIR_FALLING,
>>> + .mask_separate = BIT(IIO_EV_INFO_VALUE),
>>> + }, {
>>> + .type = IIO_EV_TYPE_THRESH,
>>> + .dir = IIO_EV_DIR_EITHER,
>>> + .mask_separate = BIT(IIO_EV_INFO_ENABLE),
>>> + },
>>> +};
>>> +
>>> +static int hi8435_get_sensing_mode(struct iio_dev *idev,
>>> + const struct iio_chan_spec *chan)
>>> +{
>>> + struct hi8435_priv *priv = iio_priv(idev);
>>> + int ret;
>>> + u8 reg;
>>> +
>>> + ret = hi8435_readb(priv, HI8435_PSEN_REG, &reg);
>>> + if (ret < 0)
>>> + return ret;
>>> +
>>> + return !!(reg & BIT(chan->channel / 8));
>>> +}
>>> +
>>> +static int hi8435_set_sensing_mode(struct iio_dev *idev,
>>> + const struct iio_chan_spec *chan,
>>> + unsigned int mode)
>>> +{
>>> + struct hi8435_priv *priv = iio_priv(idev);
>>> + int ret;
>>> + u8 reg;
>>> +
>>> + mutex_lock(&priv->lock);
>>> +
>>> + ret = hi8435_readb(priv, HI8435_PSEN_REG, &reg);
>>> + if (ret < 0) {
>>> + mutex_unlock(&priv->lock);
>>> + return ret;
>>> + }
>>> +
>>> + reg &= ~BIT(chan->channel / 8);
>>> + if (mode)
>>> + reg |= BIT(chan->channel / 8);
>>> +
>>> + ret = hi8435_writeb(priv, HI8435_PSEN_REG, reg);
>>> +
>>> + mutex_unlock(&priv->lock);
>>> +
>>> + return ret;
>>> +}
>>> +
>>> +static const char * const hi8435_sensing_modes[] = { "GND-Open",
>>> + "Supply-Open" };
>>> +
>>> +static const struct iio_enum hi8435_sensing_mode = {
>>> + .items = hi8435_sensing_modes,
>>> + .num_items = ARRAY_SIZE(hi8435_sensing_modes),
>>> + .get = hi8435_get_sensing_mode,
>>> + .set = hi8435_set_sensing_mode,
>>> +};
>>> +
>>> +static const struct iio_chan_spec_ext_info hi8435_ext_info[] = {
>>> + IIO_ENUM("sensing_mode", IIO_SEPARATE, &hi8435_sensing_mode),
>>> + {},
>>> +};
>>> +
>>> +#define HI8435_VOLTAGE_CHANNEL(num) \
>>> +{ \
>>> + .type = IIO_VOLTAGE, \
>>> + .indexed = 1, \
>>> + .channel = num, \
>>> + .info_mask_separate = BIT(IIO_CHAN_INFO_RAW), \
If we are using events, why have the raw access at all?
>>> + .info_mask_shared_by_all = BIT(IIO_CHAN_INFO_DEBOUNCE_TIME), \
>>> + .event_spec = hi8435_events, \
>>> + .num_event_specs = ARRAY_SIZE(hi8435_events), \
>>> + .ext_info = hi8435_ext_info, \
>>> + .extend_name = "comparator" \
>> I'm really not sure about this approach. Can see where you are coming
>> from. We already have an events interface supporting these things so
>> why not use it? I'm just doubtful it is an efficient option or a clean
>> interface (would like more opinions on this!)
> Sorry, I did not understand here.
> Is it about .extend_name ?
> I did add "comparator" extention to have attributes look like "in_voltage0_comparator_raw"
I'd drop it, particularly if we also drop the raw access.
>>> +}
>>> +
>>> +static const struct iio_chan_spec hi8435_channels[] = {
>>> + HI8435_VOLTAGE_CHANNEL(0),
>>> + HI8435_VOLTAGE_CHANNEL(1),
>>> + HI8435_VOLTAGE_CHANNEL(2),
>>> + HI8435_VOLTAGE_CHANNEL(3),
>>> + HI8435_VOLTAGE_CHANNEL(4),
>>> + HI8435_VOLTAGE_CHANNEL(5),
>>> + HI8435_VOLTAGE_CHANNEL(6),
>>> + HI8435_VOLTAGE_CHANNEL(7),
>>> + HI8435_VOLTAGE_CHANNEL(8),
>>> + HI8435_VOLTAGE_CHANNEL(9),
>>> + HI8435_VOLTAGE_CHANNEL(10),
>>> + HI8435_VOLTAGE_CHANNEL(11),
>>> + HI8435_VOLTAGE_CHANNEL(12),
>>> + HI8435_VOLTAGE_CHANNEL(13),
>>> + HI8435_VOLTAGE_CHANNEL(14),
>>> + HI8435_VOLTAGE_CHANNEL(15),
>>> + HI8435_VOLTAGE_CHANNEL(16),
>>> + HI8435_VOLTAGE_CHANNEL(17),
>>> + HI8435_VOLTAGE_CHANNEL(18),
>>> + HI8435_VOLTAGE_CHANNEL(19),
>>> + HI8435_VOLTAGE_CHANNEL(20),
>>> + HI8435_VOLTAGE_CHANNEL(21),
>>> + HI8435_VOLTAGE_CHANNEL(22),
>>> + HI8435_VOLTAGE_CHANNEL(23),
>>> + HI8435_VOLTAGE_CHANNEL(24),
>>> + HI8435_VOLTAGE_CHANNEL(25),
>>> + HI8435_VOLTAGE_CHANNEL(26),
>>> + HI8435_VOLTAGE_CHANNEL(27),
>>> + HI8435_VOLTAGE_CHANNEL(28),
>>> + HI8435_VOLTAGE_CHANNEL(29),
>>> + HI8435_VOLTAGE_CHANNEL(30),
>>> + HI8435_VOLTAGE_CHANNEL(31),
>>> + IIO_CHAN_SOFT_TIMESTAMP(32),
>>> +};
>>> +
>>> +static const struct iio_info hi8435_info = {
>>> + .driver_module = THIS_MODULE,
>>> + .attrs = &hi8435_attribute_group,
>>> + .read_raw = hi8435_read_raw,
>>> + .write_raw = hi8435_write_raw,
>>> + .read_event_config = &hi8435_read_event_config,
>>> + .write_event_config = hi8435_write_event_config,
>>> + .read_event_value = &hi8435_read_event_value,
>>> + .write_event_value = &hi8435_write_event_value,
>>> +};
>>> +
>>> +static void hi8435_iio_push_event(struct iio_dev *idev, unsigned int val)
>>> +{
>>> + struct hi8435_priv *priv = iio_priv(idev);
>>> + enum iio_event_direction dir;
>>> + unsigned int i;
>>> + unsigned int status = priv->event_prev_val ^ val;
>>> +
>>> + if (!status)
>>> + return;
>>> +
>>> + for_each_set_bit(i, &priv->event_scan_mask, 32) {
>>> + if (!(status & BIT(i)))
>>> + continue;
>>> +
>>> + dir = val & BIT(i) ? IIO_EV_DIR_RISING : IIO_EV_DIR_FALLING;
>>> +
>>> + iio_push_event(idev,
>>> + IIO_UNMOD_EVENT_CODE(IIO_VOLTAGE, i,
>>> + IIO_EV_TYPE_THRESH, dir),
>>> + iio_get_time_ns());
>>> + }
>>> +
>>> + priv->event_prev_val = val;
>>> +}
>>> +
>>> +static void hi8435_debounce_work(struct work_struct *work)
>>> +{
>>> + struct hi8435_priv *priv = container_of(work, struct hi8435_priv,
>>> + work.work);
>>> + struct iio_dev *idev = spi_get_drvdata(priv->spi);
>>> + u32 val;
>>> + int ret;
>>> +
>>> + ret = hi8435_readl(priv, HI8435_SO31_0_REG, &val);
>>> + if (ret < 0)
>>> + return;
>>> +
>>> + if (val == priv->debounce_val)
>>> + hi8435_iio_push_event(idev, val);
>>> + else
>>> + dev_warn(&priv->spi->dev, "filtered by software debounce");
>> Definitely not. Warning every time a standard event occurs? Not
>> a good idea. Use the debug stuff if you want a way of knowing this
>> happened, then it will silent under normal use.
> I'd say it is not standard event, but occasional/debounce event.
> Anyway I agree with you.
>
>>> +}
>>> +
>>> +static irqreturn_t hi8435_trigger_handler(int irq, void *private)
>>> +{
>>> + struct iio_poll_func *pf = private;
>>> + struct iio_dev *idev = pf->indio_dev;
>>> + struct hi8435_priv *priv = iio_priv(idev);
>>> + u32 val;
>>> + int ret;
>>> +
>>> + ret = hi8435_readl(priv, HI8435_SO31_0_REG, &val);
>>> + if (ret < 0)
>>> + goto err_read;
>>> +
>>> + if (priv->debounce_interval) {
>>> + priv->debounce_val = val;
>>> + schedule_delayed_work(&priv->work,
>>> + msecs_to_jiffies(priv->debounce_interval));
>> This is another element that makes me doubt that the event interface
>> is the way to do this. I'd much rather we pushed the debouncing out
>> to userspace where cleverer things can be done (and more adaptive things).
>> Here to keep the event flow low you have to do it in the kernel.
> Yes, it is helpful for reducing the data flow (buffer data or event data)
> Many kernel drivers support s/w event debouncing, f.e. usb cable plug-in, or other connectors.
True enough that it is common functionality. However, we try to keep
the IIO datahandling as minimal as possible and leave the job to userspace.

We aren't likely to be dealing with huge datarates here (as we are effectively
polling at maybe a few hundred Hz). Also, with the hysterisis set sensibly
seems unlikely that much bounce due to noise will ever occur. If you get
bounce that overcomes that then you probably have a wiring or logic problem!
>
>>> + } else {
>>> + hi8435_iio_push_event(idev, val);
>>> + }
>>> +
>>> +err_read:
>>> + iio_trigger_notify_done(idev->trig);
>>> +
>>> + return IRQ_HANDLED;
>>> +}
>>> +
>>> +static void hi8435_parse_dt(struct hi8435_priv *priv)
>>> +{
>>> + struct device_node *np = priv->spi->dev.of_node;
>>> + int ret;
>>> +
>>> + ret = of_get_named_gpio(np, "holt,reset-gpios", 0);
>>> + priv->reset_gpio = ret < 0 ? 0 : ret;
>> Silly question, but can gpio0 exist on a board? I suspect you
>> may need an additional variable to hold that this request failed
>> rather than using the magic value of 0.
> I will do handle this case, thank you.
Look at what Lars added on this question...
>>> +
>>> + ret = of_property_read_u32(np, "holt,debounce-interval",
>>> + &priv->debounce_interval);
>>> + if (ret)
>>> + priv->debounce_interval = 0;
>>> + if (priv->debounce_interval > HI8435_DEBOUNCE_DELAY_MAX)
>>> + priv->debounce_interval = HI8435_DEBOUNCE_DELAY_MAX;
>>> +}
>>> +
>>> +static int hi8435_probe(struct spi_device *spi)
>>> +{
>>> + struct iio_dev *idev;
>>> + struct hi8435_priv *priv;
>>> + int ret;
>>> +
>>> + idev = devm_iio_device_alloc(&spi->dev, sizeof(*priv));
>>> + if (!idev)
>>> + return -ENOMEM;
>>> +
>>> + priv = iio_priv(idev);
>>> + priv->spi = spi;
>>> +
>>> + if (spi->dev.of_node)
>>> + hi8435_parse_dt(priv);
>>> +
>>> + spi_set_drvdata(spi, idev);
>>> + mutex_init(&priv->lock);
>>> + INIT_DELAYED_WORK(&priv->work, hi8435_debounce_work);
>>> +
>>> + idev->dev.parent = &spi->dev;
>>> + idev->name = spi_get_device_id(spi)->name;
>>> + idev->modes = INDIO_DIRECT_MODE;
>>> + idev->info = &hi8435_info;
>>> + idev->channels = hi8435_channels;
>>> + idev->num_channels = ARRAY_SIZE(hi8435_channels);
>>> +
>>> + if (priv->reset_gpio) {
>>> + ret = devm_gpio_request(&spi->dev, priv->reset_gpio, idev->name);
>>> + if (!ret) {
>>> + /* chip hardware reset */
>>> + gpio_direction_output(priv->reset_gpio, 0);
>>> + udelay(5);
>>> + gpio_direction_output(priv->reset_gpio, 1);
>>> + }
>>> + } else {
>>> + /* chip software reset */
>>> + hi8435_writeb(priv, HI8435_CTRL_REG, HI8435_CTRL_SRST);
>>> + /* get out from reset state */
>>> + hi8435_writeb(priv, HI8435_CTRL_REG, 0);
>>> + }
>>> +
>>> + /* unmask all events */
>>> + priv->event_scan_mask = ~(0);
>>> + /* initialize default thresholds */
>>> + priv->threshold_lo[0] = priv->threshold_lo[1] = 2;
>>> + priv->threshold_hi[0] = priv->threshold_hi[1] = 4;
>> These seem very random. Why these values or
> The aim for thresholds initialization is that they are in undefined state after h/w reset and
> the driver allows userspace to change thresholds (high or low) separately.
>
> There is a restriction in the chip - the hysteresis can not be even.
> If the hysteresis is set to even value then it get's into lock state and not functional anymore. (I've figured this out empirically).
*nice*
> So we need to initialize thresholds (aka hysteresis + it's center) to some initial value to be able to change
> high and low thresholds separately and avoid even values for hysteresis.
>
> I will also add these comments.
Cool
>>> + hi8435_writew(priv, HI8435_GOCENHYS_REG, 0x206);
>> What is 0x206? Again, I guess a default. I'd like to see a comment
>> here saying what real world value it corresponds to.
> Yes it is default values.
>
> The 0x206 is a code for HI threshold voltage = 4V and low voltage threshold = 2V.
> The register HI8435_SOCENHYS_REG describes the voltage hysteresis of the sensor and the hysteresis center.
> The 0x206 is an encoded value for 2V-to-4V voltage threshold.
>
> I will add comments.
good.
>>> + hi8435_writew(priv, HI8435_SOCENHYS_REG, 0x206);
>>> +
>> I am as yet unconvinced that using triggers to poll events is a terribly
>> good idea. Feels rather like we are doing this due to a deficiency in
>> the buffer interface than because sending events makes sense here.
> Actually the first version of this driver was against the buffer interface.
> I did try to use your tip and switched to event interface :)
oops.
>
> Do you think I should switch to buffer interface back?
Nope. As discussed above, right now this approach is winning (just ;)
>>> + ret = iio_triggered_event_setup(idev, NULL, hi8435_trigger_handler);
>>> + if (ret)
>>> + return ret;
>>> +
>>> + ret = iio_device_register(idev);
>>> + if (ret < 0) {
>>> + dev_err(&spi->dev, "unable to register device\n");
>>> + goto unregister_triggered_event;
>>> + }
>>> +
>>> + return 0;
>>> +
>>> +unregister_triggered_event:
>>> + iio_triggered_event_cleanup(idev);
>>> + return ret;
>>> +}
>>> +
>>> +static int hi8435_remove(struct spi_device *spi)
>>> +{
>>> + struct iio_dev *idev = spi_get_drvdata(spi);
>>> + struct hi8435_priv *priv = iio_priv(idev);
>>> +
>>> + cancel_delayed_work_sync(&priv->work);
>>> + iio_device_unregister(idev);
>>> + iio_triggered_event_cleanup(idev);
>>> +
>>> + return 0;
>>> +}
>>> +
>>> +static const struct of_device_id hi8435_dt_ids[] = {
>>> + { .compatible = "holt,hi8435" },
>>> + {},
>>> +};
>>> +MODULE_DEVICE_TABLE(of, hi8435_dt_ids);
>>> +
>>> +static const struct spi_device_id hi8435_id[] = {
>>> + { "hi8435", 0},
>>> + { }
>>> +};
>>> +MODULE_DEVICE_TABLE(spi, hi8435_id);
>>> +
>>> +static struct spi_driver hi8435_driver = {
>>> + .driver = {
>>> + .name = DRV_NAME,
>>> + .of_match_table = of_match_ptr(hi8435_dt_ids),
>>> + },
>>> + .probe = hi8435_probe,
>>> + .remove = hi8435_remove,
>>> + .id_table = hi8435_id,
>>> +};
>>> +module_spi_driver(hi8435_driver);
>>> +
>>> +MODULE_LICENSE("GPL");
>>> +MODULE_AUTHOR("Vladimir Barinov");
>>> +MODULE_DESCRIPTION("HI-8435 threshold detector");
>>>
>
> --
> To unsubscribe from this list: send the line "unsubscribe linux-iio" in
> the body of a message to majordomo@xxxxxxxxxxxxxxx
> More majordomo info at http://vger.kernel.org/majordomo-info.html

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