Message-ID: <20180921132012.00007061@huawei.com>
Date:   Fri, 21 Sep 2018 13:20:12 +0100
From:   Jonathan Cameron <jonathan.cameron@...wei.com>
To:     Song Qiang <songqiang1304521@...il.com>
CC:     Phil Reid <preid@...ctromag.com.au>,
        Peter Meerwald-Stadler <pmeerw@...erw.net>, <jic23@...nel.org>,
        <lars@...afoo.de>, <linux-iio@...r.kernel.org>,
        <devicetree@...r.kernel.org>, <linux-kernel@...r.kernel.org>
Subject: Re: [PATCH] iio: magnetometer: Add support for PNI RM3100 9-axis
 magnetometer

Hi Song,

...

Please crop emails down to the relevant section when replying.  Lots of scrolling
for everyone otherwise and sometimes things get missed when doing that.

> > > > +	if (!buffer)
> > > > +		goto done;
> > > > +
> > > > +	mutex_lock(&data->lock);
> > > > +	ret = rm3100_wait_measurement(data);
> > > > +	if (ret < 0) {
> > > > +		mutex_unlock(&data->lock);
> > > > +		goto done;
> > > > +	}
> > > > +
> > > > +	for (i = 0; i < 3; i++) {
> > > > +		ret = regmap_bulk_read(regmap, RM_REG_MX2 + 3 * i,
> > > > +				buffer + 4 * i, 3);
> > > > +		if (ret < 0)
> > > > +			return ret;
> > > > +	}  
> > 
> > Wouldn't it be better to read the 3 axis with one transaction here.
> > And if required shuffle the data into the iio buffer.
> > 
> >   
> 
> Hi Phil,
> 
> That's surely something will make this code better!
> 
> But also, there is something worth discussing,
> When triggered buffer is triggered here, I should push all the data into
> userspace, and every time the buffer is enabled, the iio subsystem
> automatically computes the alignment of my data. This sensor has 3
> 24bits channels, and my original thought of data alignment should be
> like this because my machine is 32bits:
> +----------+----+----+----+----+----+----+
> | 3b(ytes) | 1b | 3b | 1b | 3b | 1b | 8b |
> +----------+----+----+----+----+----+----+
> 3 bytes data and 1 byte for alignment, last 8 bytes for the timestamp.
> Aligned to 4 bytes. 
> But the actual layout is like this:
> +----------+----+----+----+----+----+----+----+
> | 3b(ytes) | 1b | 3b | 1b | 3b | 1b | 4b | 8b |
> +----------+----+----+----+----+----+----+----+
> Soon after I debugged the code of industrial io core I found that in the
> iio_compute_scanbytes() in industialio-buffer.c, alignment was computed
> each channel a time, and was aligned with the length of next channel. (eg:
> In the last case, before the timestamp channel was appended, there are
> 12 bytes already in the buffer, to align to the timetamp which is 8 
> bytes, it injected 4 bytes to make it 16, and then append timestamp. )
> This leads to two questions:
> 
> 1. I thought computer systems align their data in the memory for least
> time access, shouldn't this always align to the bits of architecture,
> like for a 32bits machine align to 4?? I'm a little confused here.
> Jonathan must have considered something, Hoping to get some
> explanations!

A couple of reasons for this:

1. It's much nicer to have the same spacing independent of the architecture
as it saves having to have userspace code do different things depending on
what the kernel is running at, which may not be the same as userspace.

2. The size of a memory read is not the only scale that makes sense for
alignment.  In theory at least you can have fun and games like crossing of
cacheline boundaries if you do 64 bit access only at 32bits aligned on a
32 bit platform.  Won't hit all that often with typically 64 byte cachelines
but it is the sort of subtle effect you get when you don't go with
'naturally' aligned. (which is what aligning to an elements own size is
normally called IIRC).

> 
> 2. In this case I have 4 channels including the timetamp one, and I
> thought it should be aligned like the first one. But actually it's
> aligned like the second one, when I was testing the buffer, I got
> confused, because the in_magn_x_type returned 24/32, 24/32, 24/32
> and 64/64. If the iio core injected some empty bytes, shouldn't it
> notify me? Or at least change the timestamp type in scan_bytes to
> 64/96>>0?? And when I only enabled two of three channels, the data  
> alignment in the buffer looks like this:
> +----------+----+----+----+----+
> | 3b(ytes) | 1b | 3b | 1b | 8b |
> +----------+----+----+----+----+
> Notice that 4 bytes have gone due to data of two channels have already
> aligned to 8 bytes.
> Hoping to get some answers, I'm very curious and interested about this!

Nope it shouldn't do the 64/96 need to tell you because it uses
the rules of natural alignment which are easy for userspace to predict.
So in this second case there is no need to pad so it doesn't - whereas the
4 bytes are needed in the first case.

We could have explicitly output the offsets and kept things more flexible
but that would have prevented some general optimizations on the userspace
side for common cases.  If you are reading a 3 axis accelerometer for
example with timestamp then you can probably safely prebake optimized handling
for,

1B (x), 1B (y), 1B(z), 5B(pad), 8B(ts)

2B (x), 2B (y), 2B(z), 2B(pad), 8B(ts)

in reality no one ever goes higher than this because accelerometers aren't
accurate enough...

If we did anything more complex we'd need to describe to userspace in some
sense what was happening beyond the static data currently provided (when
combined with the ordering and enables) and that makes for a complex ABI
we don't need.

Also we'd have to do unaligned puts of the u64 timestamp in kernel which
is nasty.

Jonathan
> 
> yours,
> Song Qiang
> 
> > > > +	mutex_unlock(&data->lock);
> > > > +
> > > > +	iio_push_to_buffers_with_timestamp(indio_dev, buffer,
> > > > +			iio_get_time_ns(indio_dev));
> > > > +done:
> > > > +	iio_trigger_notify_done(indio_dev->trig);
> > > > +
> > > > +	return IRQ_HANDLED;
> > > > +}
> > > > +
> > > > +int rm3100_common_probe(struct device *dev, struct regmap *regmap, int irq)
> > > > +{
> > > > +	struct iio_dev *indio_dev;
> > > > +	struct rm3100_data *data;
> > > > +	int tmp;
> > > > +	int ret;
> > > > +
> > > > +	indio_dev = devm_iio_device_alloc(dev, sizeof(*data));
> > > > +	if (!indio_dev)
> > > > +		return -ENOMEM;
> > > > +
> > > > +	data = iio_priv(indio_dev);
> > > > +	dev_set_drvdata(dev, indio_dev);
> > > > +	data->dev = dev;
> > > > +	data->regmap = regmap;
> > > > +
> > > > +	mutex_init(&data->lock);
> > > > +
> > > > +	indio_dev->dev.parent = dev;
> > > > +	indio_dev->name = "rm3100";
> > > > +	indio_dev->info = &rm3100_info;
> > > > +	indio_dev->channels = rm3100_channels;
> > > > +	indio_dev->num_channels = ARRAY_SIZE(rm3100_channels);
> > > > +	indio_dev->modes = INDIO_DIRECT_MODE;
> > > > +	indio_dev->available_scan_masks = rm3100_scan_masks;
> > > > +
> > > > +	if (!irq)
> > > > +		data->use_interrupt = false;
> > > > +	else {
> > > > +		data->use_interrupt = true;
> > > > +		ret = devm_request_irq(dev,
> > > > +			irq,
> > > > +			rm3100_measurement_irq_handler,
> > > > +			IRQF_TRIGGER_RISING,
> > > > +			indio_dev->name,
> > > > +			data);
> > > > +		if (ret < 0) {
> > > > +			dev_err(dev,
> > > > +			"request irq line failed.");  
> > > 
> > > \n
> > >   
> > > > +			return -ret;
> > > > +		}
> > > > +		init_completion(&data->measuring_done);
> > > > +	}
> > > > +
> > > > +	ret = iio_triggered_buffer_setup(indio_dev, NULL,
> > > > +					rm3100_trigger_handler, NULL);
> > > > +	if (ret < 0)
> > > > +		return ret;
> > > > +
> > > > +	/* 3sec more wait time. */
> > > > +	ret = regmap_read(data->regmap, RM_REG_TMRC, &tmp);  
> > > 
> > > check ret
> > >   
> > > > +	data->conversion_time = rm3100_samp_rates[tmp-RM_TMRC_OFFSET][2] + 3000;
> > > > +
> > > > +	/* Starting all channels' conversion. */
> > > > +	ret = regmap_write(regmap, RM_REG_CMM,
> > > > +		RM_CMM_PMX | RM_CMM_PMY | RM_CMM_PMZ | RM_CMM_START);
> > > > +	if (ret < 0)
> > > > +		return ret;
> > > > +
> > > > +	return devm_iio_device_register(dev, indio_dev);
> > > > +}
> > > > +EXPORT_SYMBOL(rm3100_common_probe);
> > > > +
> > > > +int rm3100_common_remove(struct device *dev)
> > > > +{
> > > > +	struct iio_dev *indio_dev = dev_get_drvdata(dev);
> > > > +	struct rm3100_data *data = iio_priv(indio_dev);
> > > > +	struct regmap *regmap = data->regmap;
> > > > +
> > > > +	regmap_write(regmap, RM_REG_CMM, 0x00);
> > > > +
> > > > +	return 0;
> > > > +}
> > > > +EXPORT_SYMBOL(rm3100_common_remove);
> > > > +
> > > > +MODULE_AUTHOR("Song Qiang <songqiang1304521@...il.com>");
> > > > +MODULE_DESCRIPTION("PNI RM3100 9-axis magnetometer i2c driver");
> > > > +MODULE_LICENSE("GPL v2");
> > > > diff --git a/drivers/iio/magnetometer/rm3100-i2c.c b/drivers/iio/magnetometer/rm3100-i2c.c
> > > > new file mode 100644
> > > > index 000000000000..b50dc5b1b30b
> > > > --- /dev/null
> > > > +++ b/drivers/iio/magnetometer/rm3100-i2c.c
> > > > @@ -0,0 +1,66 @@
> > > > +// SPDX-License-Identifier: GPL-2.0+
> > > > +/*
> > > > + * Support for PNI RM3100 9-axis geomagnetic sensor a i2c bus.
> > > > + *
> > > > + * Copyright (C) 2018 Song Qiang <songqiang1304521@...il.com>
> > > > + *
> > > > + * User Manual available at
> > > > + * <https://www.pnicorp.com/download/rm3100-user-manual/>
> > > > + *
> > > > + * i2c slave address 0x20 + SA1 << 1 + SA0.
> > > > + */
> > > > +
> > > > +#include <linux/i2c.h>
> > > > +
> > > > +#include "rm3100.h"
> > > > +
> > > > +static const struct regmap_config rm3100_regmap_config = {
> > > > +		.reg_bits = 8,
> > > > +		.val_bits = 8,
> > > > +
> > > > +		.rd_table = &rm3100_readable_table,
> > > > +		.wr_table = &rm3100_writable_table,
> > > > +		.volatile_table = &rm3100_volatile_table,
> > > > +
> > > > +		.cache_type = REGCACHE_RBTREE,
> > > > +};
> > > > +
> > > > +static int rm3100_probe(struct i2c_client *client)
> > > > +{
> > > > +	struct regmap *regmap;
> > > > +
> > > > +	if (!i2c_check_functionality(client->adapter,
> > > > +		I2C_FUNC_SMBUS_READ_I2C_BLOCK | I2C_FUNC_SMBUS_BYTE_DATA))
> > > > +		return -EOPNOTSUPP;
> > > > +
> > > > +	regmap =  devm_regmap_init_i2c(client, &rm3100_regmap_config);
> > > > +	if (IS_ERR(regmap))
> > > > +		return PTR_ERR(regmap);
> > > > +
> > > > +	return rm3100_common_probe(&client->dev, regmap, client->irq);
> > > > +}
> > > > +
> > > > +static int rm3100_remove(struct i2c_client *client)
> > > > +{
> > > > +	return rm3100_common_remove(&client->dev);
> > > > +}
> > > > +
> > > > +static const struct of_device_id rm3100_dt_match[] = {
> > > > +	{ .compatible = "pni,rm3100-i2c", },
> > > > +	{ }
> > > > +};
> > > > +MODULE_DEVICE_TABLE(of, rm3100_dt_match);
> > > > +
> > > > +static struct i2c_driver rm3100_driver = {
> > > > +	.driver = {
> > > > +		.name = "rm3100-i2c",
> > > > +		.of_match_table = rm3100_dt_match,
> > > > +	},
> > > > +	.probe_new = rm3100_probe,
> > > > +	.remove = rm3100_remove,
> > > > +};
> > > > +module_i2c_driver(rm3100_driver);
> > > > +
> > > > +MODULE_AUTHOR("Song Qiang <songqiang1304521@...il.com>");
> > > > +MODULE_DESCRIPTION("PNI RM3100 9-axis magnetometer i2c driver");
> > > > +MODULE_LICENSE("GPL v2");
> > > > diff --git a/drivers/iio/magnetometer/rm3100-spi.c b/drivers/iio/magnetometer/rm3100-spi.c
> > > > new file mode 100644
> > > > index 000000000000..2c7dd9e3a1a2
> > > > --- /dev/null
> > > > +++ b/drivers/iio/magnetometer/rm3100-spi.c
> > > > @@ -0,0 +1,72 @@
> > > > +// SPDX-License-Identifier: GPL-2.0+
> > > > +/*
> > > > + * Support for PNI RM3100 9-axis geomagnetic sensor a spi bus.
> > > > + *
> > > > + * Copyright (C) 2018 Song Qiang <songqiang1304521@...il.com>
> > > > + *
> > > > + * User Manual available at
> > > > + * <https://www.pnicorp.com/download/rm3100-user-manual/>
> > > > + */
> > > > +
> > > > +#include <linux/spi/spi.h>
> > > > +
> > > > +#include "rm3100.h"
> > > > +
> > > > +static const struct regmap_config rm3100_regmap_config = {
> > > > +		.reg_bits = 8,
> > > > +		.val_bits = 8,
> > > > +
> > > > +		.rd_table = &rm3100_readable_table,
> > > > +		.wr_table = &rm3100_writable_table,
> > > > +		.volatile_table = &rm3100_volatile_table,
> > > > +
> > > > +		.read_flag_mask = 0x80,
> > > > +
> > > > +		.cache_type = REGCACHE_RBTREE,
> > > > +};
> > > > +
> > > > +static int rm3100_probe(struct spi_device *spi)
> > > > +{
> > > > +	struct regmap *regmap;
> > > > +	int ret;
> > > > +
> > > > +	/* Actually this device supports both mode 0 and mode 3. */
> > > > +	spi->mode = SPI_MODE_0;
> > > > +	/* data rates cannot exceeds 1Mbits. */  
> > > 
> > > exceed
> > >   
> > > > +	spi->max_speed_hz = 1000000;
> > > > +	spi->bits_per_word = 8;
> > > > +	ret = spi_setup(spi);
> > > > +	if (ret)
> > > > +		return ret;
> > > > +
> > > > +	regmap =  devm_regmap_init_spi(spi, &rm3100_regmap_config);
> > > > +	if (IS_ERR(regmap))
> > > > +		return PTR_ERR(regmap);
> > > > +
> > > > +	return rm3100_common_probe(&spi->dev, regmap, spi->irq);
> > > > +}
> > > > +
> > > > +static int rm3100_remove(struct spi_device *spi)
> > > > +{
> > > > +	return rm3100_common_remove(&spi->dev);
> > > > +}
> > > > +
> > > > +static const struct of_device_id rm3100_dt_match[] = {
> > > > +	{ .compatible = "pni,rm3100-spi", },
> > > > +	{ }
> > > > +};
> > > > +MODULE_DEVICE_TABLE(of, rm3100_dt_match);
> > > > +
> > > > +static struct spi_driver rm3100_driver = {
> > > > +		.driver = {
> > > > +				.name = "rm3100-spi",
> > > > +				.of_match_table = rm3100_dt_match,
> > > > +		},
> > > > +		.probe = rm3100_probe,
> > > > +		.remove = rm3100_remove,
> > > > +};
> > > > +module_spi_driver(rm3100_driver);
> > > > +
> > > > +MODULE_AUTHOR("Song Qiang <songqiang1304521@...il.com>");
> > > > +MODULE_DESCRIPTION("PNI RM3100 9-axis magnetometer spi driver");
> > > > +MODULE_LICENSE("GPL v2");
> > > > diff --git a/drivers/iio/magnetometer/rm3100.h b/drivers/iio/magnetometer/rm3100.h
> > > > new file mode 100644
> > > > index 000000000000..5e30bc0f5149
> > > > --- /dev/null
> > > > +++ b/drivers/iio/magnetometer/rm3100.h
> > > > @@ -0,0 +1,90 @@
> > > > +/* SPDX-License-Identifier: GPL-2.0+ */
> > > > +/*
> > > > + * Header file for PNI RM3100 driver
> > > > + *
> > > > + * Copyright (C) 2018 Song Qiang <songqiang1304521@...il.com>
> > > > + */
> > > > +
> > > > +#ifndef RM3100_CORE_H
> > > > +#define RM3100_CORE_H
> > > > +
> > > > +#include <linux/module.h>
> > > > +#include <linux/regmap.h>
> > > > +
> > > > +#define RM_REG_REV_ID		0x36
> > > > +
> > > > +/* Cycle Count Registers MSBs and LSBs. */
> > > > +#define RM_REG_CCXM		0x04
> > > > +#define RM_REG_CCXL		0x05
> > > > +#define RM_REG_CCYM		0x06
> > > > +#define RM_REG_CCYL		0x07
> > > > +#define RM_REG_CCZM		0x08
> > > > +#define RM_REG_CCZL		0x09
> > > > +
> > > > +/* Single Measurement Mode register. */
> > > > +#define RM_REG_POLL		0x00
> > > > +#define RM_POLL_PMX		BIT(4)
> > > > +#define RM_POLL_PMY		BIT(5)
> > > > +#define RM_POLL_PMZ		BIT(6)
> > > > +
> > > > +/* Continues Measurement Mode register. */
> > > > +#define RM_REG_CMM		0x01
> > > > +#define RM_CMM_START		BIT(0)
> > > > +#define RM_CMM_DRDM		BIT(2)
> > > > +#define RM_CMM_PMX		BIT(4)
> > > > +#define RM_CMM_PMY		BIT(5)
> > > > +#define RM_CMM_PMZ		BIT(6)
> > > > +
> > > > +/* TiMe Rate Configuration register. */
> > > > +#define RM_REG_TMRC		0x0B
> > > > +#define RM_TMRC_OFFSET		0x92
> > > > +
> > > > +/* Result Status register. */
> > > > +#define RM_REG_STATUS		0x34
> > > > +#define RM_STATUS_DRDY		BIT(7)
> > > > +
> > > > +/* Measurement result registers. */
> > > > +#define RM_REG_MX2		0x24
> > > > +#define RM_REG_MX1		0x25
> > > > +#define RM_REG_MX0		0x26
> > > > +#define RM_REG_MY2		0x27
> > > > +#define RM_REG_MY1		0x28
> > > > +#define RM_REG_MY0		0x29
> > > > +#define RM_REG_MZ2		0x2a
> > > > +#define RM_REG_MZ1		0x2b
> > > > +#define RM_REG_MZ0		0x2c
> > > > +
> > > > +#define RM_REG_HSHAKE		0x35
> > > > +
> > > > +#define RM_W_REG_START		RM_REG_POLL
> > > > +#define RM_W_REG_END		RM_REG_REV_ID
> > > > +#define RM_R_REG_START		RM_REG_POLL
> > > > +#define RM_R_REG_END		RM_REG_HSHAKE
> > > > +#define RM_V_REG_START		RM_REG_MX2
> > > > +#define RM_V_REG_END		RM_REG_HSHAKE
> > > > +
> > > > +/* Built-In Self Test reigister. */
> > > > +#define RM_REG_BIST		0x33
> > > > +
> > > > +struct rm3100_data {
> > > > +	struct device *dev;
> > > > +	struct regmap *regmap;
> > > > +	struct completion measuring_done;
> > > > +	bool use_interrupt;
> > > > +
> > > > +	int conversion_time;
> > > > +
> > > > +	/* To protect consistency of every measurement and sampling
> > > > +	 * frequency change operations.
> > > > +	 */
> > > > +	struct mutex lock;
> > > > +};
> > > > +
> > > > +extern const struct regmap_access_table rm3100_readable_table;
> > > > +extern const struct regmap_access_table rm3100_writable_table;
> > > > +extern const struct regmap_access_table rm3100_volatile_table;
> > > > +
> > > > +int rm3100_common_probe(struct device *dev, struct regmap *regmap, int irq);
> > > > +int rm3100_common_remove(struct device *dev);
> > > > +
> > > > +#endif /* RM3100_CORE_H */
> > > >   
> > >   
> > 
> > 
> > -- 
> > Regards
> > Phil Reid
> >   

Powered by blists - more mailing lists