Message-ID: <14d0f5fb4240a7e0c3665d4ffc128117c5515ac6.camel@gmail.com>
Date: Thu, 24 Oct 2024 15:05:39 +0200
From: Nuno Sá <noname.nuno@...il.com>
To: Angelo Dureghello <adureghello@...libre.com>
Cc: Nuno Sá <nuno.sa@...log.com>, Lars-Peter Clausen
 <lars@...afoo.de>, Michael Hennerich <Michael.Hennerich@...log.com>, 
 Jonathan Cameron <jic23@...nel.org>, Rob Herring <robh@...nel.org>,
 Krzysztof Kozlowski <krzk+dt@...nel.org>, Conor Dooley
 <conor+dt@...nel.org>, Olivier Moysan <olivier.moysan@...s.st.com>,
 linux-iio@...r.kernel.org,  devicetree@...r.kernel.org,
 linux-kernel@...r.kernel.org, dlechner@...libre.com,  Mark Brown
 <broonie@...nel.org>
Subject: Re: [PATCH v7 7/8] iio: dac: ad3552r: add high-speed platform driver

On Tue, 2024-10-22 at 18:40 +0200, Angelo Dureghello wrote:
> Hi Nuno,
> 
> On 22.10.2024 14:28, Nuno Sá wrote:
> > On Mon, 2024-10-21 at 14:40 +0200, Angelo Dureghello wrote:
> > > From: Angelo Dureghello <adureghello@...libre.com>
> > > 
> > > Add High Speed ad3552r platform driver.
> > > 
> > > The ad3552r DAC is controlled by a custom (fpga-based) DAC IP
> > > through the current AXI backend, or similar alternative IIO backend.
> > > 
> > > Compared to the existing driver (ad3552r.c), that is a simple SPI
> > > driver, this driver is coupled with a DAC IIO backend that finally
> > > controls the ad3552r by a fpga-based "QSPI+DDR" interface, to reach
> > > maximum transfer rate of 33MUPS using dma stream capabilities.
> > > 
> > > All commands involving QSPI bus read/write are delegated to the backend
> > > through the provided APIs for bus read/write.
> > > 
> > > Signed-off-by: Angelo Dureghello <adureghello@...libre.com>
> > > ---
> > >  drivers/iio/dac/Kconfig      |  14 ++
> > >  drivers/iio/dac/Makefile     |   1 +
> > >  drivers/iio/dac/ad3552r-hs.c | 547
> > > +++++++++++++++++++++++++++++++++++++++++++
> > >  drivers/iio/dac/ad3552r-hs.h |  18 ++
> > >  drivers/iio/dac/ad3552r.h    |   4 +
> > >  5 files changed, 584 insertions(+)
> > > 
> > > diff --git a/drivers/iio/dac/Kconfig b/drivers/iio/dac/Kconfig
> > > index fa091995d002..fc11698e88f2 100644
> > > --- a/drivers/iio/dac/Kconfig
> > > +++ b/drivers/iio/dac/Kconfig
> > > @@ -6,6 +6,20 @@
> > >  
> > >  menu "Digital to analog converters"
> > >  
> > > +config AD3552R_HS
> > > +	tristate "Analog Devices AD3552R DAC High Speed driver"
> > > +	select ADI_AXI_DAC
> > > +	help
> > > +	  Say yes here to build support for Analog Devices AD3552R
> > > +	  Digital to Analog Converter High Speed driver.
> > > +
> > > +          The driver requires the assistance of an IP core to operate,
> > > +          since data is streamed into target device via DMA, sent over a
> > > +	  QSPI + DDR (Double Data Rate) bus.
> > > +
> > > +	  To compile this driver as a module, choose M here: the
> > > +	  module will be called ad3552r-hs.
> > > +
> > >  config AD3552R
> > >  	tristate "Analog Devices AD3552R DAC driver"
> > >  	depends on SPI_MASTER
> > > diff --git a/drivers/iio/dac/Makefile b/drivers/iio/dac/Makefile
> > > index c92de0366238..d92e08ca93ca 100644
> > > --- a/drivers/iio/dac/Makefile
> > > +++ b/drivers/iio/dac/Makefile
> > > @@ -4,6 +4,7 @@
> > >  #
> > >  
> > >  # When adding new entries keep the list in alphabetical order
> > > +obj-$(CONFIG_AD3552R_HS) += ad3552r-hs.o ad3552r-common.o
> > >  obj-$(CONFIG_AD3552R) += ad3552r.o ad3552r-common.o
> > >  obj-$(CONFIG_AD5360) += ad5360.o
> > >  obj-$(CONFIG_AD5380) += ad5380.o
> > > diff --git a/drivers/iio/dac/ad3552r-hs.c b/drivers/iio/dac/ad3552r-hs.c
> > > new file mode 100644
> > > index 000000000000..27bdc35fdc29
> > > --- /dev/null
> > > +++ b/drivers/iio/dac/ad3552r-hs.c
> > > @@ -0,0 +1,547 @@
> > > +// SPDX-License-Identifier: GPL-2.0-only
> > > +/*
> > > + * Analog Devices AD3552R
> > > + * Digital to Analog converter driver, High Speed version
> > > + *
> > > + * Copyright 2024 Analog Devices Inc.
> > > + */
> > > +
> > > +#include <linux/bitfield.h>
> > > +#include <linux/delay.h>
> > > +#include <linux/gpio/consumer.h>
> > > +#include <linux/iio/backend.h>
> > > +#include <linux/iio/buffer.h>
> > > +#include <linux/mod_devicetable.h>
> > > +#include <linux/platform_device.h>
> > > +#include <linux/property.h>
> > > +#include <linux/units.h>
> > > +
> > > +#include "ad3552r.h"
> > > +#include "ad3552r-hs.h"
> > > +
> > > +struct ad3552r_hs_state {
> > > +	const struct ad3552r_model_data *model_data;
> > > +	struct gpio_desc *reset_gpio;
> > > +	struct device *dev;
> > > +	struct iio_backend *back;
> > > +	bool single_channel;
> > > +	struct ad3552r_ch_data ch_data[AD3552R_MAX_CH];
> > > +	struct ad3552r_hs_platform_data *data;
> > > +};
> > > +
> > > +static int ad3552r_qspi_update_reg_bits(struct ad3552r_hs_state *st,
> > > +					u32 reg, u32 mask, u32 val,
> > > +					size_t xfer_size)
> > > +{
> > > +	u32 rval;
> > > +	int ret;
> > > +
> > > +	ret = st->data->bus_reg_read(st->back, reg, &rval, xfer_size);
> > > +	if (ret)
> > > +		return ret;
> > > +
> > > +	rval = (rval & ~mask) | val;
> > > +
> > > +	return st->data->bus_reg_write(st->back, reg, rval, xfer_size);
> > > +}
> > > +
> > > +static int ad3552r_hs_read_raw(struct iio_dev *indio_dev,
> > > +			       struct iio_chan_spec const *chan,
> > > +			       int *val, int *val2, long mask)
> > > +{
> > > +	struct ad3552r_hs_state *st = iio_priv(indio_dev);
> > > +	int ret;
> > > +	int ch = chan->channel;
> > > +
> > > +	switch (mask) {
> > > +	case IIO_CHAN_INFO_SAMP_FREQ: {
> > > +		int sclk;
> > > +
> > > +		ret = iio_backend_read_raw(st->back, chan, &sclk, 0,
> > > +					   IIO_CHAN_INFO_FREQUENCY);
> > > +		if (ret != IIO_VAL_INT)
> > > +			return -EINVAL;
> > > +
> > 
> > I just saw you had some questions on v6 that everyone failed to see. See my
> > reply to David here:
> > 
> > https://lore.kernel.org/linux-iio/61cf3072af74a8b2951c948ddc2383ba1e55954d.camel@gmail.com/
> > 
> > It should be easy and it's something that makes sense (at least to me :))
> > 
> 
> I understood that we would improve things later in case.
> 
> Could we maybe stay with IIO_CHAN_INFO_FREQUENCY ? It doesn't seems to me
> so out of scope. Sorry but i am trying to finalize someway this job,
> so i am trying to conatain changes now at v7, if code is not really 
> totally wrong.

I think you're trying to rush in the series. I can understand your frustration but
believe me that v7 (or v8) is not so bad :).

David already raised concerns about using IIO_CHAN_INFO_FREQUENCY. I'm also not a fan
of it and gave you another option that should be trivial and makes sense (given that
bus_read and write are already being done through the platform_data interface). So
no, I don't think we're going to accept "is not really totally wrong.". IOW, We want
it to be totally right - if such a thing exists :).

> 
> > > +		/*
> > > +		 * Using 4 lanes (QSPI), then using 2 as DDR mode is
> > > +		 * considered always on (considering buffering mode always).
> > > +		 */
> > > +		*val = DIV_ROUND_CLOSEST(sclk * 4 * 2,
> > > +					 chan->scan_type.realbits);
> > > +
> > > +		return IIO_VAL_INT;
> > > +	}
> > > +	case IIO_CHAN_INFO_RAW:
> > > +		ret = st->data->bus_reg_read(st->back,
> > > +				AD3552R_REG_ADDR_CH_DAC_16B(chan->channel),
> > > +				val, 2);
> > > +		if (ret)
> > > +			return ret;
> > > +
> > > +		return IIO_VAL_INT;
> > > +	case IIO_CHAN_INFO_SCALE:
> > > +		*val = st->ch_data[ch].scale_int;
> > > +		*val2 = st->ch_data[ch].scale_dec;
> > > +		return IIO_VAL_INT_PLUS_MICRO;
> > > +	case IIO_CHAN_INFO_OFFSET:
> > > +		*val = st->ch_data[ch].offset_int;
> > > +		*val2 = st->ch_data[ch].offset_dec;
> > > +		return IIO_VAL_INT_PLUS_MICRO;
> > > +	default:
> > > +		return -EINVAL;
> > > +	}
> > > +}
> > > +
> > > +static int ad3552r_hs_write_raw(struct iio_dev *indio_dev,
> > > +				struct iio_chan_spec const *chan,
> > > +				int val, int val2, long mask)
> > > +{
> > > +	struct ad3552r_hs_state *st = iio_priv(indio_dev);
> > > +
> > > +	switch (mask) {
> > > +	case IIO_CHAN_INFO_RAW:
> > > +		iio_device_claim_direct_scoped(return -EBUSY, indio_dev) {
> > > +			return st->data->bus_reg_write(st->back,
> > > +				    AD3552R_REG_ADDR_CH_DAC_16B(chan-
> > > > channel),
> > > +				    val, 2);
> > > +		}
> > 
> > Maybe we'll get the new stuff in time for this :)
> > 
> > ...
> > 
> > > +
> > > +static int ad3552r_hs_reset(struct ad3552r_hs_state *st)
> > > +{
> > > +	int ret;
> > > +
> > > +	/*
> > > +	 * Using inverted "active-high" logic here, since ad3552r classic-spi
> > > +	 * fdt node (and driver) is using the same logic.
> > > +	 */
> > > +
> > 
> > I don't understand this. This is a new device with a different compatible. Why
> > keeping the wrong logic? AFAICT, there's nothing in the bindings about the pin
> > polarity.
> > 
> 
> ad3552r.c uses same compatible (adi,ad3552r), and in the code it implements 
> this same inverted logic. So i thought to use the same logic.
> I can anyway change to the correct active-low logic for this driver, 
> but would honestly not enter in fixing old code now at v7. 
> Happy to do such fix on ad3552r.c later on.

Ok, bad example from me with the compatible. The point is this is a different device.
It's a platform device while the other one is a spi device. So why doing it wrong in
here? Not saying to change the other device logic, just not doing it deliberately
wrong in a new device.

For the old device, we can't likely change it as we could break current users who
just adapted their DTs to conform to the driver logic.

> 
> > > +	st->reset_gpio = devm_gpiod_get_optional(st->dev,
> > > +						 "reset", GPIOD_OUT_LOW);
> > > +	if (IS_ERR(st->reset_gpio))
> > > +		return PTR_ERR(st->reset_gpio);
> > > +
> > > +	if (st->reset_gpio) {
> > > +		fsleep(10);
> > > +		gpiod_set_value_cansleep(st->reset_gpio, 1);
> > > +	} else {
> > > +		ret = ad3552r_qspi_update_reg_bits(st,
> > > +					AD3552R_REG_ADDR_INTERFACE_CONFIG_A,
> > > +					AD3552R_MASK_SOFTWARE_RESET,
> > > +					AD3552R_MASK_SOFTWARE_RESET, 1);
> > > +		if (ret)
> > > +			return ret;
> > > +	}
> > > +	msleep(100);
> > > +
> > > +	return 0;
> > > +}
> > > +
> > > +static int ad3552r_hs_scratch_pad_test(struct ad3552r_hs_state *st)
> > > +{
> > > +	int ret, val;
> > > +
> > > +	ret = st->data->bus_reg_write(st->back, AD3552R_REG_ADDR_SCRATCH_PAD,
> > > +				      AD3552R_SCRATCH_PAD_TEST_VAL1, 1);
> > > +	if (ret)
> > > +		return ret;
> > > +
> > > +	ret = st->data->bus_reg_read(st->back, AD3552R_REG_ADDR_SCRATCH_PAD,
> > > +				     &val, 1);
> > > +	if (ret)
> > > +		return ret;
> > > +
> > > +	if (val != AD3552R_SCRATCH_PAD_TEST_VAL1)
> > > +		return dev_err_probe(st->dev, -EIO,
> > > +			"SCRATCH_PAD_TEST mismatch. Expected 0x%x, Read
> > > 0x%x\n",
> > > +			AD3552R_SCRATCH_PAD_TEST_VAL1, val);
> > > +
> > > +	ret = st->data->bus_reg_write(st->back, AD3552R_REG_ADDR_SCRATCH_PAD,
> > > +				      AD3552R_SCRATCH_PAD_TEST_VAL2, 1);
> > > +	if (ret)
> > > +		return ret;
> > > +
> > > +	ret = st->data->bus_reg_read(st->back, AD3552R_REG_ADDR_SCRATCH_PAD,
> > > +				     &val, 1);
> > > +	if (ret)
> > > +		return ret;
> > > +
> > > +	if (val != AD3552R_SCRATCH_PAD_TEST_VAL2)
> > > +		return dev_err_probe(st->dev, -EIO,
> > > +			"SCRATCH_PAD_TEST mismatch. Expected 0x%x, Read
> > > 0x%x\n",
> > > +			AD3552R_SCRATCH_PAD_TEST_VAL2, val);
> > > +
> > > +	return 0;
> > > +}
> > > +
> > > +static int ad3552r_hs_setup_custom_gain(struct ad3552r_hs_state *st,
> > > +					int ch, u16 gain, u16 offset)
> > > +{
> > > +	int ret;
> > > +
> > > +	ret = st->data->bus_reg_write(st->back,
> > > AD3552R_REG_ADDR_CH_OFFSET(ch),
> > > +				      offset, 1);
> > > +	if (ret)
> > > +		return dev_err_probe(st->dev, ret, "Error writing
> > > register\n");
> > > +
> > > +	ret = st->data->bus_reg_write(st->back, AD3552R_REG_ADDR_CH_GAIN(ch),
> > > +				      gain, 1);
> > > +	if (ret)
> > > +		return dev_err_probe(st->dev, ret, "Error writing
> > > register\n");
> > > +
> > > +	return 0;
> > 
> > nit: Not a big fan of these logs on read/write registers functions... Also seems
> > that you're not being consistent (either you have them or not). FWIW, I would
> > simplify and drop them. That would allow to do
> > 
> > return st->data->bus_reg_write(st->back, AD3552R_REG_ADDR_CH_GAIN(ch), gain, 1);
> > 
> 
> Used dev_err_probe on quite all probe functions.
> I don't see nothing really wrong on this codem except maybe a more meaningful
> message.

No, you're not being consistent. You have another calls (example: st->data-
>bus_reg_rea()) where no log is being given. 

> 
> > > +}
> > > +
> > > +static int ad3552r_hs_setup(struct ad3552r_hs_state *st)
> > > +{
> > > +	s16 goffs;
> > > +	u16 id;
> > > +	u16 gain = 0, offset = 0;
> > > +	u32 ch, val, range;
> > > +	int ret;
> > > +
> > > +	ret = ad3552r_hs_reset(st);
> > > +	if (ret)
> > > +		return ret;
> > > +
> > > +	ret = iio_backend_ddr_disable(st->back);
> > > +	if (ret)
> > > +		return ret;
> > > +
> > > +	ret = ad3552r_hs_scratch_pad_test(st);
> > > +	if (ret)
> > > +		return ret;
> > > +
> > > +	ret = st->data->bus_reg_read(st->back, AD3552R_REG_ADDR_PRODUCT_ID_L,
> > > +				     &val, 1);
> > > +	if (ret)
> > > +		return ret;
> > > +
> > > +	id = val;
> > > +
> > > +	ret = st->data->bus_reg_read(st->back, AD3552R_REG_ADDR_PRODUCT_ID_H,
> > > +				     &val, 1);
> > > +	if (ret)
> > > +		return ret;
> > > +
> > > +	id |= val << 8;
> > > +	if (id != st->model_data->chip_id)
> > > +		dev_info(st->dev, "Chip ID error. Expected 0x%x, Read
> > > 0x%x\n",
> > > +			 AD3552R_ID, id);
> > > +
> > > +	ret = st->data->bus_reg_write(st->back,
> > > +				      AD3552R_REG_ADDR_SH_REFERENCE_CONFIG,
> > > +				      0, 1);
> > > +	if (ret)
> > > +		return ret;
> > > +
> > > +	ret = st->data->bus_reg_write(st->back,
> > > +				AD3552R_REG_ADDR_TRANSFER_REGISTER,
> > > +				FIELD_PREP(AD3552R_MASK_MULTI_IO_MODE,
> > > +					   AD3552R_QUAD_SPI) |
> > > +				AD3552R_MASK_STREAM_LENGTH_KEEP_VALUE, 1);
> > > +	if (ret)
> > > +		return ret;
> > > +
> > > +	ret = iio_backend_data_source_set(st->back, 0, IIO_BACKEND_EXTERNAL);
> > > +	if (ret)
> > > +		return ret;
> > > +
> > > +	ret = iio_backend_data_source_set(st->back, 1, IIO_BACKEND_EXTERNAL);
> > > +	if (ret)
> > > +		return ret;
> > > +
> > > +	ret = ad3552r_get_ref_voltage(st->dev, &val);
> > > +	if (ret < 0)
> > > +		return ret;
> > > +
> > > +	val = ret;
> > > +
> > > +	ret = ad3552r_qspi_update_reg_bits(st,
> > > +				AD3552R_REG_ADDR_SH_REFERENCE_CONFIG,
> > > +				AD3552R_MASK_REFERENCE_VOLTAGE_SEL,
> > > +				val, 1);
> > > +	if (ret)
> > > +		return ret;
> > > +
> > > +	ret = ad3552r_get_drive_strength(st->dev, &val);
> > > +	if (!ret) {
> > > +		ret = ad3552r_qspi_update_reg_bits(st,
> > > +					AD3552R_REG_ADDR_INTERFACE_CONFIG_D,
> > > +					AD3552R_MASK_SDO_DRIVE_STRENGTH,
> > > +					val, 1);
> > > +		if (ret)
> > > +			return ret;
> > > +	}
> > > +
> > > +	device_for_each_child_node_scoped(st->dev, child) {
> > > +		ret = fwnode_property_read_u32(child, "reg", &ch);
> > > +		if (ret)
> > > +			return dev_err_probe(st->dev, ret,
> > > +					     "reg property missing\n");
> > > +
> > > +		ret = ad3552r_get_output_range(st->dev, st->model_data,
> > > child,
> > > +					       &range);
> > > +		if (!ret) {
> > > +			st->ch_data[ch].range = range;
> > > +
> > > +			ret = ad3552r_hs_set_output_range(st, ch, range);
> > > +			if (ret)
> > > +				return ret;
> > > +
> > > +		} else if (ret == -ENOENT) {
> > > +			ret = ad3552r_get_custom_gain(st->dev, child,
> > > +						&st->ch_data[ch].p,
> > > +						&st->ch_data[ch].n,
> > > +						&st->ch_data[ch].rfb,
> > > +						&st-
> > > > ch_data[ch].gain_offset);
> > > +			if (ret)
> > > +				return ret;
> > > +
> > > +			gain = ad3552r_calc_custom_gain(st->ch_data[ch].p,
> > > +						st->ch_data[ch].n,
> > > +						st->ch_data[ch].gain_offset);
> > > +			offset = abs(goffs);
> > > +
> > > +			st->ch_data[ch].range_override = 1;
> > > +
> > > +			ret = ad3552r_hs_setup_custom_gain(st, ch, gain,
> > > +							   offset);
> > > +			if (ret)
> > > +				return ret;
> > > +		} else {
> > > +			return ret;
> > > +		}
> > 
> > Just personal preference... I think this would be neater:
> > if (ret && ret != ENOENT)
> > 	return ret;
> > if (ret == -ENOENT) {
> > 	...
> > } else {
> > 	...
> > }
> > 
> > Advantage is that it also handles errors first (which is the typical pattern)
> 
> I tested this code, would not change possibly now at this stage,
> unless another version should be sent.

Not sure if we'll need another but personally I cannot ack this one as it stands...
sorry.

>  
> > 
> > > +
> > > +		ad3552r_calc_gain_and_offset(&st->ch_data[ch], st-
> > > > model_data);
> > > +	}
> > > +
> > > +	return 0;
> > > +}
> > > +
> > > +static const struct iio_buffer_setup_ops ad3552r_hs_buffer_setup_ops = {
> > > +	.postenable = ad3552r_hs_buffer_postenable,
> > > +	.predisable = ad3552r_hs_buffer_predisable,
> > > +};
> > > +
> > > +#define AD3552R_CHANNEL(ch) { \
> > > +	.type = IIO_VOLTAGE, \
> > > +	.info_mask_separate = BIT(IIO_CHAN_INFO_RAW) | \
> > > +			      BIT(IIO_CHAN_INFO_SAMP_FREQ) | \
> > > +			      BIT(IIO_CHAN_INFO_SCALE) | \
> > > +			      BIT(IIO_CHAN_INFO_OFFSET), \
> > > +	.output = 1, \
> > > +	.indexed = 1, \
> > > +	.channel = (ch), \
> > > +	.scan_index = (ch), \
> > > +	.scan_type = { \
> > > +		.sign = 'u', \
> > > +		.realbits = 16, \
> > > +		.storagebits = 16, \
> > > +		.endianness = IIO_BE, \
> > > +	} \
> > > +}
> > > +
> > > +static const struct iio_chan_spec ad3552r_hs_channels[] = {
> > > +	AD3552R_CHANNEL(0),
> > > +	AD3552R_CHANNEL(1),
> > > +};
> > > +
> > > +static const struct iio_info ad3552r_hs_info = {
> > > +	.read_raw = &ad3552r_hs_read_raw,
> > > +	.write_raw = &ad3552r_hs_write_raw,
> > > +};
> > > +
> > > +static int ad3552r_hs_probe(struct platform_device *pdev)
> > > +{
> > > +	struct ad3552r_hs_state *st;
> > > +	struct iio_dev *indio_dev;
> > > +	int ret;
> > > +
> > > +	indio_dev = devm_iio_device_alloc(&pdev->dev, sizeof(*st));
> > > +	if (!indio_dev)
> > > +		return -ENOMEM;
> > > +
> > > +	st = iio_priv(indio_dev);
> > > +	st->dev = &pdev->dev;
> > > +
> > > +	st->data = pdev->dev.platform_data;
> > 
> > dev_get_platdata()
> > 
> 
> pdev->dev.platform_data seems correct to me, used in a lot of places
> in the driver framework. Can we stay with it ?
> 

It is correct but if we an helper, why not using it? It may be used in a lot of
places just because the helper was added afterwards...

- Nuno Sá
> 

Powered by blists - more mailing lists