Received-SPF: pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
Date:   Mon, 13 Apr 2020 17:47:10 +0100
From:   Jonathan Cameron <jic23@jic23.retrosnub.co.uk>
To:     Heiko Stuebner <heiko@sntech.de>
Cc:     knaack.h@gmx.de, lars@metafoo.de, pmeerw@pmeerw.net,
        linux-iio@vger.kernel.org, linux-kernel@vger.kernel.org,
        linux-rockchip@lists.infradead.org, xxm@rock-chips.com,
        kever.yang@rock-chips.com,
        Heiko Stuebner <heiko.stuebner@theobroma-systems.com>
Subject: Re: [PATCH v3] iio: adc: rockchip_saradc: Add support iio buffers
Message-ID: <20200413174710.06224388@archlinux>
In-Reply-To: <20200413174434.55b2941a@archlinux>
References: <20200412224251.2919182-1-heiko@sntech.de>
        <20200413174434.55b2941a@archlinux>
MIME-Version: 1.0
Content-Type: text/plain; charset=US-ASCII
Content-Transfer-Encoding: 7bit
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

On Mon, 13 Apr 2020 17:44:34 +0100
Jonathan Cameron <jic23@kernel.org> wrote:

> On Mon, 13 Apr 2020 00:42:51 +0200
> Heiko Stuebner <heiko@sntech.de> wrote:
> 
> > From: Simon Xue <xxm@rock-chips.com>
> > 
> > Add the ability to also support access via (triggered) buffers
> > next to the existing direct mode.
> > 
> > Device in question is the Odroid Go Advance that connects a joystick
> > to two of the saradc channels for X and Y axis and the new (and still
> > pending) adc joystick driver of course wants to use triggered buffers
> > from the iio subsystem.
> > 
> > Signed-off-by: Simon Xue <xxm@rock-chips.com>
> > [some simplifications and added commit description]
> > Signed-off-by: Heiko Stuebner <heiko.stuebner@theobroma-systems.com>  
> 
> Comments inline.
> 
> The issue with mixing managed and unmanaged allocations needs tidying up.
> Sorry if I missed that one before; I probably didn't open up the current
> driver to sanity check it :(
> 
> Jonathan
> 
> > ---
> > changes in v3:
> > - split buffer struct into values and timestamp area similar to dln2-adc
> >   and make sure timestamp gets 8-byte aligned - ALIGN uses 4 as it aligns
> >   u16 elements not bytes - hopefully I got it right this time ;-)
> > changes in v2:
> > - use devm_iio_triggered_buffer_setup
> > - calculate data array size from channel number (curtesy of at91-sama5d2_adc)
> > 
> >  drivers/iio/adc/Kconfig           |   2 +
> >  drivers/iio/adc/rockchip_saradc.c | 146 ++++++++++++++++++++++--------
> >  2 files changed, 112 insertions(+), 36 deletions(-)
> > 
> > diff --git a/drivers/iio/adc/Kconfig b/drivers/iio/adc/Kconfig
> > index 12bb8b7ca1ff..8d2dd60614c6 100644
> > --- a/drivers/iio/adc/Kconfig
> > +++ b/drivers/iio/adc/Kconfig
> > @@ -809,6 +809,8 @@ config ROCKCHIP_SARADC
> >  	tristate "Rockchip SARADC driver"
> >  	depends on ARCH_ROCKCHIP || (ARM && COMPILE_TEST)
> >  	depends on RESET_CONTROLLER
> > +	select IIO_BUFFER
> > +	select IIO_TRIGGERED_BUFFER
> >  	help
> >  	  Say yes here to build support for the SARADC found in SoCs from
> >  	  Rockchip.
> > diff --git a/drivers/iio/adc/rockchip_saradc.c b/drivers/iio/adc/rockchip_saradc.c
> > index 582ba047c4a6..0713363a4b43 100644
> > --- a/drivers/iio/adc/rockchip_saradc.c
> > +++ b/drivers/iio/adc/rockchip_saradc.c
> > @@ -15,7 +15,11 @@
> >  #include <linux/delay.h>
> >  #include <linux/reset.h>
> >  #include <linux/regulator/consumer.h>
> > +#include <linux/iio/buffer.h>
> >  #include <linux/iio/iio.h>
> > +#include <linux/iio/trigger.h>
> > +#include <linux/iio/trigger_consumer.h>
> > +#include <linux/iio/triggered_buffer.h>
> >  
> >  #define SARADC_DATA			0x00
> >  
> > @@ -32,9 +36,12 @@
> >  #define SARADC_DLY_PU_SOC_MASK		0x3f
> >  
> >  #define SARADC_TIMEOUT			msecs_to_jiffies(100)
> > +#define SARADC_MAX_CHANNELS		6
> > +
> > +/* buffer elements are u16, timestamp needs to be 8-byte aligned */
> > +#define SARADC_BUFFER_NUM_U16	ALIGN(SARADC_MAX_CHANNELS, 4)  
> I may be going crazy but I think that will get you the 'start' of the
> timestamp, not the length including it.
> 
> We should be seeing 24 bytes here = 12 u16s.  Sanity check the value.
> 
> Running through the stack of defines.
> #define ALIGN(x, a)		__ALIGN_KERNEL((x), (a))
> #define __ALIGN_KERNEL(x, a)		__ALIGN_KERNEL_MASK(x, (typeof(x))(a) - 1)
> #define __ALIGN_KERNEL_MASK(x, mask)	(((x) + (mask)) & ~(mask))
> 
> ALIGN(6, 4) == __ALIGN_KERNEL(6, 4)
>             == __ALIGN_KERNEL_MASK(6, 3)
>             == (((6 + 3) & ~3) 
> which is 9 with the bottom two bits masked or b1001 & b1100 = 8 not 12
> 
> So I think you are looking for
> ALIGN(SARADC_MAX_CHANNELS + sizeof(u64) / sizeof(u16), 4)
> which will be ((10 + 3) & ~3) b1101 & b1100 = 12

Doh. I see you then added the 64 bit as a separate element in
a structure.  Note with that you will get the padding for
free anyway as this should be standard C alignment.

That's actually quite a nice way of ensuring we get the right
length!

Thanks,

Jonathan

>  
> >  
> >  struct rockchip_saradc_data {
> > -	int				num_bits;
> >  	const struct iio_chan_spec	*channels;
> >  	int				num_channels;
> >  	unsigned long			clk_rate;
> > @@ -49,8 +56,37 @@ struct rockchip_saradc {
> >  	struct reset_control	*reset;
> >  	const struct rockchip_saradc_data *data;
> >  	u16			last_val;
> > +	const struct iio_chan_spec *last_chan;
> >  };
> >  
> > +static void rockchip_saradc_power_down(struct rockchip_saradc *info)
> > +{
> > +	/* Clear irq & power down adc */
> > +	writel_relaxed(0, info->regs + SARADC_CTRL);
> > +}
> > +
> > +static int rockchip_saradc_conversion(struct rockchip_saradc *info,
> > +				   struct iio_chan_spec const *chan)
> > +{
> > +	reinit_completion(&info->completion);
> > +
> > +	/* 8 clock periods as delay between power up and start cmd */
> > +	writel_relaxed(8, info->regs + SARADC_DLY_PU_SOC);
> > +
> > +	info->last_chan = chan;
> > +
> > +	/* Select the channel to be used and trigger conversion */
> > +	writel(SARADC_CTRL_POWER_CTRL
> > +			| (chan->channel & SARADC_CTRL_CHN_MASK)
> > +			| SARADC_CTRL_IRQ_ENABLE,
> > +		   info->regs + SARADC_CTRL);
> > +
> > +	if (!wait_for_completion_timeout(&info->completion, SARADC_TIMEOUT))
> > +		return -ETIMEDOUT;
> > +
> > +	return 0;
> > +}
> > +
> >  static int rockchip_saradc_read_raw(struct iio_dev *indio_dev,
> >  				    struct iio_chan_spec const *chan,
> >  				    int *val, int *val2, long mask)
> > @@ -62,24 +98,12 @@ static int rockchip_saradc_read_raw(struct iio_dev *indio_dev,
> >  	case IIO_CHAN_INFO_RAW:
> >  		mutex_lock(&indio_dev->mlock);
> >  
> > -		reinit_completion(&info->completion);
> > -
> > -		/* 8 clock periods as delay between power up and start cmd */
> > -		writel_relaxed(8, info->regs + SARADC_DLY_PU_SOC);
> > -
> > -		/* Select the channel to be used and trigger conversion */
> > -		writel(SARADC_CTRL_POWER_CTRL
> > -				| (chan->channel & SARADC_CTRL_CHN_MASK)
> > -				| SARADC_CTRL_IRQ_ENABLE,
> > -		       info->regs + SARADC_CTRL);
> > -
> > -		if (!wait_for_completion_timeout(&info->completion,
> > -						 SARADC_TIMEOUT)) {
> > -			writel_relaxed(0, info->regs + SARADC_CTRL);
> > +		ret = rockchip_saradc_conversion(info, chan);
> > +		if (ret) {
> > +			rockchip_saradc_power_down(info);
> >  			mutex_unlock(&indio_dev->mlock);
> > -			return -ETIMEDOUT;
> > +			return ret;
> >  		}
> > -
> >  		*val = info->last_val;
> >  		mutex_unlock(&indio_dev->mlock);
> >  		return IIO_VAL_INT;
> > @@ -91,7 +115,7 @@ static int rockchip_saradc_read_raw(struct iio_dev *indio_dev,
> >  		}
> >  
> >  		*val = ret / 1000;
> > -		*val2 = info->data->num_bits;
> > +		*val2 = chan->scan_type.realbits;
> >  		return IIO_VAL_FRACTIONAL_LOG2;
> >  	default:
> >  		return -EINVAL;
> > @@ -104,10 +128,9 @@ static irqreturn_t rockchip_saradc_isr(int irq, void *dev_id)
> >  
> >  	/* Read value */
> >  	info->last_val = readl_relaxed(info->regs + SARADC_DATA);
> > -	info->last_val &= GENMASK(info->data->num_bits - 1, 0);
> > +	info->last_val &= GENMASK(info->last_chan->scan_type.realbits - 1, 0);
> >  
> > -	/* Clear irq & power down adc */
> > -	writel_relaxed(0, info->regs + SARADC_CTRL);
> > +	rockchip_saradc_power_down(info);
> >  
> >  	complete(&info->completion);
> >  
> > @@ -118,51 +141,55 @@ static const struct iio_info rockchip_saradc_iio_info = {
> >  	.read_raw = rockchip_saradc_read_raw,
> >  };
> >  
> > -#define ADC_CHANNEL(_index, _id) {				\
> > +#define ADC_CHANNEL(_index, _id, _res) {			\
> >  	.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),	\
> >  	.datasheet_name = _id,					\
> > +	.scan_index = _index,					\
> > +	.scan_type = {						\
> > +		.sign = 'u',					\
> > +		.realbits = _res,				\
> > +		.storagebits = 16,				\
> > +		.endianness = IIO_LE,				\
> > +	},							\
> >  }
> >  
> >  static const struct iio_chan_spec rockchip_saradc_iio_channels[] = {
> > -	ADC_CHANNEL(0, "adc0"),
> > -	ADC_CHANNEL(1, "adc1"),
> > -	ADC_CHANNEL(2, "adc2"),
> > +	ADC_CHANNEL(0, "adc0", 10),
> > +	ADC_CHANNEL(1, "adc1", 10),
> > +	ADC_CHANNEL(2, "adc2", 10),
> >  };
> >  
> >  static const struct rockchip_saradc_data saradc_data = {
> > -	.num_bits = 10,
> >  	.channels = rockchip_saradc_iio_channels,
> >  	.num_channels = ARRAY_SIZE(rockchip_saradc_iio_channels),
> >  	.clk_rate = 1000000,
> >  };
> >  
> >  static const struct iio_chan_spec rockchip_rk3066_tsadc_iio_channels[] = {
> > -	ADC_CHANNEL(0, "adc0"),
> > -	ADC_CHANNEL(1, "adc1"),
> > +	ADC_CHANNEL(0, "adc0", 12),
> > +	ADC_CHANNEL(1, "adc1", 12),
> >  };
> >  
> >  static const struct rockchip_saradc_data rk3066_tsadc_data = {
> > -	.num_bits = 12,
> >  	.channels = rockchip_rk3066_tsadc_iio_channels,
> >  	.num_channels = ARRAY_SIZE(rockchip_rk3066_tsadc_iio_channels),
> >  	.clk_rate = 50000,
> >  };
> >  
> >  static const struct iio_chan_spec rockchip_rk3399_saradc_iio_channels[] = {
> > -	ADC_CHANNEL(0, "adc0"),
> > -	ADC_CHANNEL(1, "adc1"),
> > -	ADC_CHANNEL(2, "adc2"),
> > -	ADC_CHANNEL(3, "adc3"),
> > -	ADC_CHANNEL(4, "adc4"),
> > -	ADC_CHANNEL(5, "adc5"),
> > +	ADC_CHANNEL(0, "adc0", 10),
> > +	ADC_CHANNEL(1, "adc1", 10),
> > +	ADC_CHANNEL(2, "adc2", 10),
> > +	ADC_CHANNEL(3, "adc3", 10),
> > +	ADC_CHANNEL(4, "adc4", 10),
> > +	ADC_CHANNEL(5, "adc5", 10),
> >  };
> >  
> >  static const struct rockchip_saradc_data rk3399_saradc_data = {
> > -	.num_bits = 10,
> >  	.channels = rockchip_rk3399_saradc_iio_channels,
> >  	.num_channels = ARRAY_SIZE(rockchip_rk3399_saradc_iio_channels),
> >  	.clk_rate = 1000000,
> > @@ -193,6 +220,42 @@ static void rockchip_saradc_reset_controller(struct reset_control *reset)
> >  	reset_control_deassert(reset);
> >  }
> >  
> > +static irqreturn_t rockchip_saradc_trigger_handler(int irq, void *p)
> > +{
> > +	struct iio_poll_func *pf = p;
> > +	struct iio_dev *i_dev = pf->indio_dev;
> > +	struct rockchip_saradc *info = iio_priv(i_dev);
> > +	struct {
> > +		u16 values[SARADC_BUFFER_NUM_U16];
> > +		int64_t timestamp;
> > +	} data;
> > +	int ret;
> > +	int i, j = 0;
> > +
> > +	mutex_lock(&i_dev->mlock);
> > +
> > +	for_each_set_bit(i, i_dev->active_scan_mask, i_dev->masklength) {
> > +		const struct iio_chan_spec *chan = &i_dev->channels[i];
> > +
> > +		ret = rockchip_saradc_conversion(info, chan);
> > +		if (ret) {
> > +			rockchip_saradc_power_down(info);
> > +			goto out;
> > +		}
> > +
> > +		data.values[j] = info->last_val;
> > +		j++;
> > +	}
> > +
> > +	iio_push_to_buffers_with_timestamp(i_dev, &data, iio_get_time_ns(i_dev));
> > +out:
> > +	mutex_unlock(&i_dev->mlock);
> > +
> > +	iio_trigger_notify_done(i_dev->trig);
> > +
> > +	return IRQ_HANDLED;
> > +}
> > +
> >  static int rockchip_saradc_probe(struct platform_device *pdev)
> >  {
> >  	struct rockchip_saradc *info = NULL;
> > @@ -221,6 +284,11 @@ static int rockchip_saradc_probe(struct platform_device *pdev)
> >  
> >  	info->data = match->data;
> >  
> > +	if (info->data->num_channels > SARADC_MAX_CHANNELS) {
> > +		dev_err(&pdev->dev, "max channels exceeded");
> > +		return -EINVAL;  
> 
> How can that happen?  Bug in the addition of a new device type?
> If it's just paranoia against future code, perhaps add a comment to
> say that.
> 
> > +	}
> > +
> >  	mem = platform_get_resource(pdev, IORESOURCE_MEM, 0);
> >  	info->regs = devm_ioremap_resource(&pdev->dev, mem);
> >  	if (IS_ERR(info->regs))
> > @@ -315,6 +383,12 @@ static int rockchip_saradc_probe(struct platform_device *pdev)
> >  	indio_dev->channels = info->data->channels;
> >  	indio_dev->num_channels = info->data->num_channels;
> >  
> > +	ret = devm_iio_triggered_buffer_setup(&indio_dev->dev, indio_dev, NULL,
> > +					      rockchip_saradc_trigger_handler,
> > +					      NULL);
> > +	if (ret)
> > +		goto err_clk;
> > +  
> 
> Please avoid mixing an matching between device managed an unmanaged interfaces.
> It means the driver is not 'obviously correct' and hence harder to review.
> 
> Two choices here.  Either use devm_add_action_or_reset to automatically
> disable each clock + regulator in the managed release path, drop all the error
> handling and remove (note this should be a precursor patch), or use
> iio_triggered_buffer_setup and manually call iio_triggered_buffer_cleanup
> in the right place in the remove function.
>  
> >  	ret = iio_device_register(indio_dev);
> >  	if (ret)
> >  		goto err_clk;  
>