Re: [PATCH] iio: add support of the max1027

From: Jonathan Cameron
Date: Sat May 10 2014 - 06:35:10 EST


On 10/05/14 01:43, Hartmut Knaack wrote:
Philippe Reynes schrieb:
This driver add partial support of the
maxim 1027/1029/1031. Differential mode is not
supported.

It was tested on armadeus apf27 board.

Signed-off-by: Philippe Reynes <tremyfr@xxxxxxxx>

I've added a few additional comments to Hartmut's thorough review.
Mostly cases where the driver could be simplified a bit.

Nice to see these spi parts finally getting support (I have one
in a box somewhere I was meaning to solder up a good 4 years
ago ;(
---
.../devicetree/bindings/iio/adc/max1027-adc.txt | 19 +
drivers/staging/iio/adc/Kconfig | 10 +
drivers/staging/iio/adc/Makefile | 1 +
drivers/staging/iio/adc/max1027.c | 647 ++++++++++++++++++++
4 files changed, 677 insertions(+), 0 deletions(-)
create mode 100644 Documentation/devicetree/bindings/iio/adc/max1027-adc.txt
create mode 100644 drivers/staging/iio/adc/max1027.c

diff --git a/Documentation/devicetree/bindings/iio/adc/max1027-adc.txt b/Documentation/devicetree/bindings/iio/adc/max1027-adc.txt
new file mode 100644
index 0000000..8daafe4
--- /dev/null
+++ b/Documentation/devicetree/bindings/iio/adc/max1027-adc.txt
@@ -0,0 +1,19 @@
+* Maxim 1027/1029/1031 Analog to Digital Converter (ADC)
+
+Required properties:
+ - compatible: Should be "maxim,max1027" or "maxim,max1029" or "maxim,max1031"
+ - reg: Should contain the ADC SPI address
+ - gpios: gpio for end of conversion
+
+Example:
+adc@0 {
+ compatible = "maxim,max1027";
+ pinctrl-names = "default";
+ pinctrl-0 = <&pinctrl_max1027>;
+ /* SPI mode = 0 */
+ spi-cpol = <0>;
+ spi-cpha = <0>;
+ spi-max-frequency = <1000000>;
+ reg = <0>;
+ gpios = <&gpio5 15 0>;
+};
diff --git a/drivers/staging/iio/adc/Kconfig b/drivers/staging/iio/adc/Kconfig
index 3633298..472504b 100644
--- a/drivers/staging/iio/adc/Kconfig
+++ b/drivers/staging/iio/adc/Kconfig
@@ -112,6 +112,16 @@ config LPC32XX_ADC
activate only one via device tree selection. Provides direct access
via sysfs.

+config MAX1027
+ tristate "Maxim max1027 ADC driver"
+ depends on SPI
+ select IIO_BUFFER
+ select IIO_TRIGGERED_BUFFER
+ help
+ Say yes here to build support for many Maxim spi analog to digital
+ converters (ADC). (max1027, max1029, max2031) Provides
Something went wrong here with the models in parenthesis. Also, we are in the ADC submenu, so whoever gets to this point should know what ADC stands for. Maybe better write something like "... build support for Maxim SPI ADC models max1027, max1029 and max1031."
+ direct access via sysfs and buffered data via the iio dev interface.
+
config MXS_LRADC
tristate "Freescale i.MX23/i.MX28 LRADC"
depends on ARCH_MXS || COMPILE_TEST
diff --git a/drivers/staging/iio/adc/Makefile b/drivers/staging/iio/adc/Makefile
index 3e9fb14..22fbd0c 100644
--- a/drivers/staging/iio/adc/Makefile
+++ b/drivers/staging/iio/adc/Makefile
@@ -18,5 +18,6 @@ obj-$(CONFIG_AD7816) += ad7816.o
obj-$(CONFIG_AD7192) += ad7192.o
obj-$(CONFIG_AD7280) += ad7280a.o
obj-$(CONFIG_LPC32XX_ADC) += lpc32xx_adc.o
+obj-$(CONFIG_MAX1027) += max1027.o
obj-$(CONFIG_MXS_LRADC) += mxs-lradc.o
obj-$(CONFIG_SPEAR_ADC) += spear_adc.o
diff --git a/drivers/staging/iio/adc/max1027.c b/drivers/staging/iio/adc/max1027.c
new file mode 100644
index 0000000..a50978f
--- /dev/null
+++ b/drivers/staging/iio/adc/max1027.c
@@ -0,0 +1,647 @@
+ /*
+ * iio/adc/max1027.c
+ * Copyright (C) 2014 Philippe Reynes
+ *
+ * based on linux/drivers/iio/ad7923.c
+ * Copyright 2011 Analog Devices Inc (from AD7923 Driver)
+ * Copyright 2012 CS Systemes d'Information
+ *
+ * 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.
+ *
+ * max1027.c
+ *
+ * Partial support for max1027 and similar chips.
+ *
+ */
Better remove the blank comment line and add a blank line after the comment block.
+#include <linux/kernel.h>
+#include <linux/spi/spi.h>
+#include <linux/module.h>
+#include <linux/interrupt.h>
+#include <linux/mutex.h>
+#include <linux/of.h>
+#include <linux/of_gpio.h>
+#include <linux/platform_device.h>
+#include <linux/delay.h>
+#include <linux/gpio.h>
+
+#include <linux/iio/iio.h>
+#include <linux/iio/buffer.h>
+#include <linux/iio/trigger.h>
+#include <linux/iio/trigger_consumer.h>
+#include <linux/iio/triggered_buffer.h>
+
+enum max1027_id {
+ max1027,
+ max1029,
+ max1031,
+};
+
You could be brave and take a look at the other similar Maxim devices.
For example based on a really quick look, the max1026, max1028 and max1030 are
very nearly the same, but just running on 3V rather than 5V.
There are also max12xx parts that are probably much the same but 12bit.
(I'm being lazy today and haven't actually compared datasheets - going
on the basis of maxims usual part numbering.
If there spi parts are anything like their i2c ones there will be dozens
of very similar parts!

+static const struct spi_device_id max1027_id[] = {
+ {"max1027", max1027},
+ {"max1029", max1029},
+ {"max1031", max1031},
+ {}
+};
+MODULE_DEVICE_TABLE(spi, max1027_id);
+
+#ifdef CONFIG_OF
+static const struct of_device_id max1027_adc_dt_ids[] = {
+ { .compatible = "maxim,max1027" },
+ { .compatible = "maxim,max1029" },
+ { .compatible = "maxim,max1031" },
+ {},
+};
+MODULE_DEVICE_TABLE(of, max1027_adc_dt_ids);
+#endif
+
+
One blank line should be enough.
+#define MAX1027_V_CHAN(index) \
+ { \
+ .type = IIO_VOLTAGE, \
+ .indexed = 1, \
+ .channel = index, \
+ .info_mask_separate = BIT(IIO_CHAN_INFO_RAW), \
+ .info_mask_shared_by_type = BIT(IIO_CHAN_INFO_SCALE), \
+ .address = index, \
+ .scan_index = index+1, \
+ .scan_type = { \
+ .sign = 'u', \
+ .realbits = 10, \
+ .storagebits = 16, \
+ .shift = 2, \
+ .endianness = IIO_BE, \
+ }, \
+ }
+
+#define MAX1027_T_CHAN(index) \
+ { \
+ .type = IIO_TEMP, \
+ .indexed = 0, \
+ .channel = index, \
+ .info_mask_separate = BIT(IIO_CHAN_INFO_RAW), \
+ .info_mask_shared_by_type = BIT(IIO_CHAN_INFO_SCALE), \
+ .address = index, \
+ .scan_index = index, \
+ .scan_type = { \
+ .sign = 'u', \
+ .realbits = 12, \
+ .storagebits = 16, \
+ .endianness = IIO_BE, \
+ }, \
+ }
+
+static const struct iio_chan_spec max1027_channels[] = {
+ MAX1027_T_CHAN(0),
+ MAX1027_V_CHAN(0),
+ MAX1027_V_CHAN(1),
+ MAX1027_V_CHAN(2),
+ MAX1027_V_CHAN(3),
+ MAX1027_V_CHAN(4),
+ MAX1027_V_CHAN(5),
+ MAX1027_V_CHAN(6),
+ MAX1027_V_CHAN(7)
+};
+
+static const struct iio_chan_spec max1029_channels[] = {
+ MAX1027_T_CHAN(0),
+ MAX1027_V_CHAN(0),
+ MAX1027_V_CHAN(1),
+ MAX1027_V_CHAN(2),
+ MAX1027_V_CHAN(3),
+ MAX1027_V_CHAN(4),
+ MAX1027_V_CHAN(5),
+ MAX1027_V_CHAN(6),
+ MAX1027_V_CHAN(7),
+ MAX1027_V_CHAN(8),
+ MAX1027_V_CHAN(9),
+ MAX1027_V_CHAN(10),
+ MAX1027_V_CHAN(11)
+};
+
+static const struct iio_chan_spec max1031_channels[] = {
+ MAX1027_T_CHAN(0),
+ MAX1027_V_CHAN(0),
+ MAX1027_V_CHAN(1),
+ MAX1027_V_CHAN(2),
+ MAX1027_V_CHAN(3),
+ MAX1027_V_CHAN(4),
+ MAX1027_V_CHAN(5),
+ MAX1027_V_CHAN(6),
+ MAX1027_V_CHAN(7),
+ MAX1027_V_CHAN(8),
+ MAX1027_V_CHAN(9),
+ MAX1027_V_CHAN(10),
+ MAX1027_V_CHAN(11),
+ MAX1027_V_CHAN(12),
+ MAX1027_V_CHAN(13),
+ MAX1027_V_CHAN(14),
+ MAX1027_V_CHAN(15)
+};
+
+struct max1027_chip_info {
+ const struct iio_chan_spec *channels;
+ unsigned int num_channels;
+};
+
+static struct max1027_chip_info max1027_chip_info[] = {
That should be "static const struct". And choosing a less confusing name, like max1027_chip_info_tbl[] would be better.

Also, right now the info structure probably just makes things more complex.
I'd be tempted to have a switch statement in the probe that simply
puts these two elements in place directly.

Might be worth having such a structure later as more stuff gets supported,
but better to leave introducing it for then.
+ [max1027] = {
+ .channels = max1027_channels,
+ .num_channels = ARRAY_SIZE(max1027_channels),
+ },
+ [max1029] = {
+ .channels = max1029_channels,
+ .num_channels = ARRAY_SIZE(max1029_channels),
+ },
+ [max1031] = {
+ .channels = max1031_channels,
+ .num_channels = ARRAY_SIZE(max1031_channels),
+ },
+};
+
+struct max1027_state {
+ struct max1027_chip_info *info;
+ struct spi_device *spi;
+ struct spi_transfer read_xfer;
+ struct spi_message read_msg;
+ struct spi_transfer write_xfer;
+ struct spi_message write_msg;
+ struct iio_trigger *trig;
+ unsigned int gpio_eoc;
+ unsigned short *buffer;
+
+ /*
+ * DMA (thus cache coherency maintenance) requires the
+ * transfer buffers to live in their own cache lines.
+ * The buffer needs to be large enough to hold two samples (4 bytes) and
+ * the naturally aligned timestamp (8 bytes).
+ */
+ struct {
+ __be16 sample[2];
+ s64 timestamp;
Why is the timestamp in here? I guess it can be aliased, but you aren't
reading it from anything that requires cache aligning... Actually on closer
inspection you aren't using this at all!

+ } data ____cacheline_aligned;
+};
+
+static int max1027_write_register(struct max1027_state *st, unsigned char reg)
+{
+ int ret;
Is it me, or is this just a complex way of expressing little endian?
+ __be16 value = cpu_to_be16(reg << 8);
+
+ pr_debug("%s(st = 0x%p, reg = 0x%02x)\n", __func__, st, reg);
+
+ st->data.sample[0] = value;
+ st->write_xfer.len = 1;
+ ret = spi_sync(st->spi, &st->write_msg);
+ if (ret < 0)
+ pr_err("cant write register\n");
"can't"
+
+ return ret;
+}
+

I vaguely wonder if the slight improvement in efficiency in reallocating
spi structures isn't outweighed by additional code complexity. Perhaps
you'd be better off using spi_read and spi_write?

+static int max1027_read_value(struct max1027_state *st, int len)
+{
+ int ret;
+
+ pr_debug("%s(st = 0x%p, len = %d)\n", __func__, st, len);
+
+ st->read_xfer.len = len;
+ ret = spi_sync(st->spi, &st->read_msg);
+ if (ret < 0)
+ pr_err("cant read value\n");
"can't"
+
+ return ret;
+}
+
+static int max1027_read_single_value(struct iio_dev *idev,
+ struct iio_chan_spec const *chan,
+ int *val, int *val2, long mask)
Common convention is naming instances of "struct iio_dev" *indio_dev. Also, aligning parameters with opening parenthesis has more beauty.
+{
+ int ret;
+ unsigned char reg;
+ struct max1027_state *st = iio_priv(idev);
+
+ if (iio_buffer_enabled(idev)) {
+ dev_warn(&idev->dev, "trigger mode already enable");
"...enabled\n");
+ return -EIO;
+ }
+
+ /*
+ * Start acquisition on conversion register write
+ * b7-b6 = 1 : setup register
+ * b5-b4 = 2 : clock mode (internal clock, start on conv write)
+ * b3-b2 = 2 : reference mode (no wake-up delay)
+ * b1-b0 = 0 : no unipolar/bipolar configuration
+ */
+ ret = max1027_write_register(st, 0x68);
The better way is to define those registers at top of the file and just put them together here using the definition names.
Absolutely agree!
+ if (ret < 0) {
+ dev_err(&idev->dev, "Failed to configure setup register\n");
+ return -EIO;
return ret?
+ }
+
+ /*
+ * b7 = 1 : conversion register
+ * b6-b3 = 0 : channel
+ * b2-b1 = 3 : no scan
+ * b0 = (chan->type == IIO_TEMP) : temp
+ */
+ reg = (1 << 7) | (chan->channel << 3) | (3 << 1);
+ if (chan->type == IIO_TEMP)
+ reg |= 1;
Same here. In the end, it should look like that:
reg = MAX1027_CONV_REG | MAX1027_CHAN(chan->channel) | MAX1027_NOSCAN | !!(chan->type == IIO_TEMP);
+ ret = max1027_write_register(st, reg);
+ if (ret < 0)
+ return ret;
+
+ /*
+ * For an unknown reason, when we use the mode "10" (write
+ * conversion register), the it doesn't occur every time.
Pardon me?
+ * So we just wait 1 ms.
+ */
+ mdelay(1);
+
+ /*
+ * Read result
+ */
+ if (chan->type == IIO_TEMP)
+ ret = max1027_read_value(st, 4);
+ else
+ ret = max1027_read_value(st, 2);
This could just be a one-liner:
ret = max1027_read_value(st, (chan->type == IIO_TEMP) ? 4 : 2);
Fine as is though for readablity reasons.. (up to you ;)
+ if (ret < 0)
+ return ret;
+
+ *val = be16_to_cpu(st->data.sample[0]);
+
+ return IIO_VAL_INT;
+}
+
+static int max1027_read_raw(struct iio_dev *idev,
+ struct iio_chan_spec const *chan,
+ int *val, int *val2, long mask)
+{
+ int ret = 0;
+
+ switch (mask) {
+ case IIO_CHAN_INFO_RAW:
+ ret = max1027_read_single_value(idev, chan, val, val2, mask);
+ break;
+ case IIO_CHAN_INFO_SCALE:
+ switch (chan->type) {
+ case IIO_TEMP:
+ *val = 1;
+ *val2 = 8;
+ ret = IIO_VAL_FRACTIONAL;
+ break;
+ case IIO_VOLTAGE:
+ *val = 1;
+ *val2 = 1;
+ ret = IIO_VAL_FRACTIONAL;
+ break;
You need to divide the reference voltage (val) by the ADC resolution (val2) and return IIO_VAL_FRACTIONAL_LOG2. So, for internal reference (as seems to be used), val = 2500; val2 = 10. The better way is to use already defined variables, and also handle external references.
+ default:
+ ret = -EINVAL;
+ break;
+ };
+ break;
+ default:
+ ret = -EINVAL;
+ break;
+ };
+
+ return ret;
+}
+
+static int max1027_update_scan_mode(struct iio_dev *idev,
+ const unsigned long *scan_mask)
+{

Now this is where we'd normally expect buffer allocation to occur and any
configuration of the spi transfers etc (if needed) - not sure they are here.

+ pr_debug("%s(idev=0x%p, scan_mask=0x%p)\n",
+ __func__, idev, scan_mask);
+ return 0;
+}
+
+static int max1027_debugfs_reg_access(struct iio_dev *idev,
+ unsigned reg, unsigned writeval,
+ unsigned *readval)
+{
+ struct max1027_state *st = iio_priv(idev);
+
Having a debug statement in the debugfs functions does seema bit over the
top.
+ pr_debug("%s(idev=0x%p, reg=%u, writeval=0x%x, readval=0x%p)\n",
+ __func__, idev, reg, writeval, readval);
+
+ if (readval != NULL)
+ return -EINVAL;
+
+ return max1027_write_register(st, writeval);
+}
+
+static int max1027_set_trigger_state(struct iio_trigger *trig, bool state)
+{
+ struct iio_dev *idev = iio_trigger_get_drvdata(trig);
+ struct max1027_state *st = iio_priv(idev);
+ int ret;
+
+ if (state) {
+ st->buffer = kmalloc(idev->scan_bytes, GFP_KERNEL);
Is this safe, or should be checked first, that st->buffer has not been allocated before (same for kfree(st->buffer) in the section below)?

This also seems a somewhat unusual place to be allocating your main
buffer... Would normally expect that to be a function of the buffer
callbacks rather than the trigger one.

This will allocate space even if (strangely) the device using the trigger
isn't this adc.
+ if (st->buffer == NULL)
+ return -ENOMEM;
+
+ /*
+ * Disable averaging
+ * b7-b5 = 1 : averaging register
+ * b4 = 0 : disable averaging
+ * b3-b2 = 0 : not used (averaging is disable)
+ * b1-b0 = 0 : not used (scan mode = 00)
+ */
+ ret = max1027_write_register(st, 0x20);
+ if (ret < 0)
+ return ret;
+
+ /*
+ * Start acquisition on cnvst falling edge
+ * b7-b6 = 1 : setup register
+ * b5-b4 = 0 : clock mode (internal clock, start on cnvst)
+ * b3-b2 = 2 : reference mode (no wake-up delay)
+ * b1-b0 = 0 : no unipolar/bipolar configuration
+ */
+ ret = max1027_write_register(st, 0x48);
+ if (ret < 0)
+ return ret;
+
+ /*
+ * we scan all the channel (0 to max)
+ * b7 = 1 : conversion register
+ * b6-b3 = 0 : channel
+ * b2-b1 = 1 : scan N to max channel
+ * b0 = 1 : temp
+ */
+ ret = max1027_write_register(st, 0x83);
+ if (ret < 0)
+ return ret;
+ } else {
+ kfree(st->buffer);
+
+ /*
+ * Disable averaging
+ * b7-b5 = 1 : averaging register
+ * b4 = 0 : disable averaging
+ * b3-b2 = 0 : not used (averaging is disable)
+ * b1-b0 = 0 : not used (scan mode = 00)
+ */
+ ret = max1027_write_register(st, 0x20);
+ if (ret < 0)
+ return ret;
+
+ /*
+ * Start acquisition on conversion register write
+ * b7-b6 = 1 : setup register
+ * b5-b4 = 2 : clock mode (internal clock, start on conv write)
+ * b3-b2 = 2 : reference mode (no wake-up delay)
+ * b1-b0 = 0 : no unipolar/bipolar configuration
+ */
+ ret = max1027_write_register(st, 0x68);
+ if (ret < 0)
+ return ret;
+ }
+
+ return 0;
+}
+
+static int max1027_try_reenable(struct iio_trigger *trig)
+{
Kill off these unused functions.
+ pr_debug("%s(trig=0x%p)\n", __func__, trig);
+ return 0;
+}
+
+static int max1027_validate_device(struct iio_trigger *trig,
+ struct iio_dev *idev)
+{
This isn't doing anything, so get rid of it please.
+ pr_debug("%s(trig=0x%p, idev=0x%p)\n", __func__, trig, idev);
+ return 0;
+}
+
+static irqreturn_t max1027_interrupt_handler(int irq, void *private)
+{
+ struct iio_dev *idev = (struct iio_dev *)private;
+
+ pr_debug("%s(irq=%d, private=0x%p)\n", __func__, irq, private);
+
If the buffer isn't enabled, wouldn't that be a bug?
+ if (iio_buffer_enabled(idev))
+ iio_trigger_poll(idev->trig, iio_get_time_ns());
+
+ return IRQ_HANDLED;
+}
+
+static irqreturn_t max1027_trigger_handler(int irq, void *private)
+{
+ struct iio_poll_func *pf = (struct iio_poll_func *)private;
+ struct iio_dev *idev = pf->indio_dev;
+ struct max1027_state *st = iio_priv(idev);
+ int i, n = 0;
+
This lot is probably excessive debugging for a driver that is ready
for submission. I know it won't do anything without debugging enabled
but is sure is uggly ;)
+ pr_debug("%s(irq=%d, private=0x%p)\n", __func__, irq, private);
+ pr_debug("%s: scan_bytes = %d\n",
+ __func__, idev->scan_bytes);
+ pr_debug("%s: scan_index_timestamp = %d\n",
+ __func__, idev->scan_index_timestamp);
+ pr_debug("%s: masklength = %d\n",
+ __func__, idev->masklength);
+
+ /* fill buffer with selected channel */
+ for (i = 0; i < idev->masklength; i++) {
+ max1027_read_value(st, 2);

You are rerolling the buffer demuxing code that the core contains.
Just put everything in the buffer every time and let that code
do it's magic. Note you'll need to specify available_scan_masks
to make that work though. For an advanced case of this see the
max1363 driver, or for a simpler one perhaps the mag3310.

The other advantage of this is that buffer then becomes a fixed length
(remember to allow for the 8 byte timestamp on the end) and hence
can just be placed directly in your state structure or
perhaps even inline in this function (as the only place it is used).

This reminds me, I should take a look to see if any other drivers
are still doing this (used to be common till we needed the demux code
to handle multiple simultaneous buffer outputs).

+ if (test_bit(i, idev->active_scan_mask)) {
+ st->buffer[n] = st->data.sample[0];
+ n++;
This could be a one-liner: st->buffer[n++] = st->data.sample[0];
Would it be necessary to check for n < ARRAY_SIZE(st->buffer[]), or is it already safe?

If not safe then there is a nasty bug hiding somewhere! This is particularly
true as the buffer also has to have space for the timestamp on the end
in the following call.
+ }
+ }
+
+ iio_push_to_buffers_with_timestamp(idev, st->buffer, iio_get_time_ns());
+
+ iio_trigger_notify_done(idev->trig);
+
+ return IRQ_HANDLED;
+}
+
+static const struct iio_trigger_ops max1027_trigger_ops = {
+ .owner = THIS_MODULE,
+ .set_trigger_state = &max1027_set_trigger_state,
+ .try_reenable = &max1027_try_reenable,
+ .validate_device = &max1027_validate_device
+};
+
+static const struct iio_info max1027_info = {
+ .driver_module = THIS_MODULE,
+ .read_raw = &max1027_read_raw,
+ .update_scan_mode = &max1027_update_scan_mode,
+ .debugfs_reg_access = &max1027_debugfs_reg_access,
+};
+
+static int max1027_probe(struct spi_device *spi)
+{
+ int err;
+ struct iio_dev *idev;
+ struct max1027_state *st;
+
+ pr_debug("%s: probe(spi = 0x%p)\n", __func__, spi);
+
+ /* allocate iio device */
+ idev = devm_iio_device_alloc(&spi->dev, sizeof(struct max1027_state));
+ if (!idev) {
+ pr_err("can't allocate iio device\n");
+ return -ENOMEM;
+ }
+
+ spi_set_drvdata(spi, idev);
+
+ st = iio_priv(idev);
+ st->spi = spi;
+ st->buffer = NULL;
+ st->info = &max1027_chip_info[spi_get_device_id(spi)->driver_data];
+
+ /* get the eoc gpio */
+ st->gpio_eoc = of_get_gpio_flags(spi->dev.of_node, 0, NULL);
+ if (st->gpio_eoc < 0) {
+ dev_err(&idev->dev, "cant get gpio eoc\n");
"can't"
+ goto fail_gpio_get;
This would result in returning err, which has not yet been set. But since we use devm_iio_device_alloc(), it is not necessary to free idev. So, the error code can be returned right here.
+ }
+
+ /* request gpio */
+ err = devm_gpio_request_one(&spi->dev, st->gpio_eoc,
+ GPIOF_IN, "max1027-eoc");
+ if (err) {
+ dev_err(&idev->dev, "cant request gpio eoc\n");
"can't"
+ goto fail_gpio_request;
Also here: directly return err.
+ }
+
+ spi->irq = gpio_to_irq(st->gpio_eoc);
+
+ /* request irq */
+ err = devm_request_threaded_irq(&spi->dev, spi->irq,
+ max1027_interrupt_handler,
+ NULL,
+ IRQF_TRIGGER_FALLING,
+ spi->dev.driver->name, idev);
+ if (err) {
+ dev_err(&idev->dev, "Failed to allocate IRQ.\n");
+ goto fail_req_irq;
+ }
+
+ /* initialiaze spi message */
"initialize"
+ st->write_xfer.tx_buf = &st->data.sample[0];
+ st->write_xfer.len = 1;
+ spi_message_init(&st->write_msg);
+ spi_message_add_tail(&st->write_xfer, &st->write_msg);

Given how often we end up with this sort of setup, I wonder if it is
worth adding an inline function to spi.h to get rid of what is effectively
boiler plate.
static inline void
spi_write_setup(const void *buf, size_t len, struct spi_message *mess,
struct spi_transfer *trans)
{
}

Then this becomes
spi_write_setup(&st->data.sample[0], 1, st->write_msg, st->write_xfer);

Probably better to consider this separately rather than with this driver
though. There is spi_init_with_transfers which will save a line of boilerplate.
e.g.

st->write_xfer.tx_buf = &st->data.sample[0];
st->write_xfer.len = 1;
spi_message_init_with_transfers(&st->write_msg, &st->write_xfer, 1);

+
+ st->read_xfer.rx_buf = &st->data.sample[0];
+ st->read_xfer.len = 4;
+ spi_message_init(&st->read_msg);
+ spi_message_add_tail(&st->read_xfer, &st->read_msg);
+
+ idev->name = spi_get_device_id(spi)->name;
+ idev->dev.parent = &spi->dev;
+ idev->info = &max1027_info;
+ idev->modes = INDIO_DIRECT_MODE;
+ idev->channels = st->info->channels;
+ idev->num_channels = st->info->num_channels;
+
+ /* Allocate buffer */
+ err = iio_triggered_buffer_setup(idev, &iio_pollfunc_store_time,
+ &max1027_trigger_handler, NULL);
+ if (err < 0) {
+ dev_err(&idev->dev, "Failed to setup buffer\n");
+ goto fail_buffer_setup;
+ }
+
+ /* Allocate trigger */
+ st->trig = devm_iio_trigger_alloc(&spi->dev, "%s-trigger",
+ idev->name);
+ if (st->trig == NULL) {
+ err = -ENOMEM;
+ dev_err(&idev->dev, "Failed to allocate iio trigger\n");
+ goto fail_trigger_alloc;
+ }
+
+ /* configure and register trigger */
+ st->trig->ops = &max1027_trigger_ops;
+ st->trig->dev.parent = &spi->dev;
+ iio_trigger_set_drvdata(st->trig, idev);
+ iio_trigger_register(st->trig);
+
+ err = iio_device_register(idev);
+ if (err < 0) {
+ dev_err(&idev->dev, "Failed to register iio device\n");
+ goto fail_dev_register;
+ }
The following setup should be done before the iio_device_register.
The device register will expose the interfaces to userspace.
From that point on you might get race conditions vs the below.
There are very few valid reasons for not having it last in any
probe function.
+
+ /*
+ * Disable averaging
+ * b7-b5 = 1 : averaging register
+ * b4 = 0 : disable averaging
+ * b3-b2 = 0 : not used (averaging is disable)
+ * b1-b0 = 0 : not used (scan mode = 00)
+ */
+ err = max1027_write_register(st, 0x20);
+ if (err < 0) {
+ dev_err(&idev->dev, "Failed to configure averaging register\n");
+ goto fail_dev_register;
+ }
+
+ /*
+ * Start acquisition on conversion register write
+ * b7-b6 = 1 : setup register
+ * b5-b4 = 2 : clock mode (internal clock, start on conv write)
+ * b3-b2 = 2 : reference mode (no wake-up delay)
+ * b1-b0 = 0 : no unipolar/bipolar configuration
+ */
+ err = max1027_write_register(st, 0x68);
+ if (err < 0) {
+ dev_err(&idev->dev, "Failed to configure setup register\n");
+ goto fail_dev_register;
+ }
+
+ return 0;
+
+fail_dev_register:
+ devm_iio_trigger_free(&idev->dev, st->trig);
+fail_trigger_alloc:
+ iio_triggered_buffer_cleanup(idev);
+fail_buffer_setup:
+ devm_free_irq(&idev->dev, spi->irq, idev);
+fail_req_irq:
+ devm_gpio_free(&idev->dev, st->gpio_eoc);
+fail_gpio_get:
+fail_gpio_request:
+ devm_iio_device_free(&spi->dev, idev);
the devm_ routines automatically clean in case of a failure.
+
+ return err;
+}
+
+static int max1027_remove(struct spi_device *spi)
+{
+ struct iio_dev *idev = spi_get_drvdata(spi);
+ struct max1027_state *st = iio_priv(idev);
+
+ pr_debug("%s: remove(spi = 0x%p)\n", __func__, spi);
+
+ iio_device_register(idev);
+ devm_iio_trigger_free(&idev->dev, st->trig);
+ iio_triggered_buffer_cleanup(idev);
+ devm_free_irq(&idev->dev, spi->irq, idev);
+ devm_gpio_free(&idev->dev, st->gpio_eoc);
+ devm_iio_device_free(&spi->dev, idev);
+
devm_ routines clean automatically.
+ return 0;
+}
+
+static struct spi_driver max1027_driver = {
+ .driver = {
+ .name = "max1027",
+ .owner = THIS_MODULE,
+ },
+ .probe = max1027_probe,
+ .remove = max1027_remove,
+ .id_table = max1027_id,
+};
+module_spi_driver(max1027_driver);
+
+MODULE_AUTHOR("Philippe Reynes <tremyfr@xxxxxxxx>");
+MODULE_DESCRIPTION("MAX1027/MAX1029/MAX1031 ADC");
+MODULE_LICENSE("GPL v2");

--
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/