Message-ID: <20180310175246.7e591c11@archlinux>
Date:   Sat, 10 Mar 2018 17:52:46 +0000
From:   Jonathan Cameron <jic23@...nel.org>
To:     Rodrigo Siqueira <rodrigosiqueiramelo@...il.com>
Cc:     Lars-Peter Clausen <lars@...afoo.de>,
        Michael Hennerich <Michael.Hennerich@...log.com>,
        Hartmut Knaack <knaack.h@....de>,
        Peter Meerwald-Stadler <pmeerw@...erw.net>,
        Greg Kroah-Hartman <gregkh@...uxfoundation.org>,
        Graff Yang <graff.yang@...il.com>, daniel.baluta@....com,
        linux-iio@...r.kernel.org, devel@...verdev.osuosl.org,
        linux-kernel@...r.kernel.org
Subject: Re: [PATCH 4/4] staging:iio:ad2s1210: Add comments/documentation

On Fri, 9 Mar 2018 20:46:40 -0300
Rodrigo Siqueira <rodrigosiqueiramelo@...il.com> wrote:

> The original code of AD2S1210 does not have documentation for structs
> and register configurations; this difficult the code comprehension. This
> patch adds structs documentation, briefly comments some register
> settings and acronyms, and adds little explanations of some calculation
> found in the code.
> 
> Signed-off-by: Rodrigo Siqueira <rodrigosiqueiramelo@...il.com>
Various comments inline.

Only a few of them are about you actual patch - mostly more general.

I'd look at renaming all those defines to be more consistent.  There
is no association between bits of a register and the register at the
moment which will make the code rather error prone.

Note this is going to be a difficult driver to get out of staging.
There is quite a bit to do and we don't currently have anyone who
has test hardware as far as I know.  So brave move ;)

Thanks,

Jonathan

> ---
>  drivers/staging/iio/resolver/ad2s1210.c | 32 ++++++++++++++++++++++++++++++++
>  drivers/staging/iio/resolver/ad2s1210.h |  9 ++++++++-
>  2 files changed, 40 insertions(+), 1 deletion(-)
> 
> diff --git a/drivers/staging/iio/resolver/ad2s1210.c b/drivers/staging/iio/resolver/ad2s1210.c
> index ac13b99bd9cb..9bb8fd782f5a 100644
> --- a/drivers/staging/iio/resolver/ad2s1210.c
> +++ b/drivers/staging/iio/resolver/ad2s1210.c
> @@ -24,8 +24,10 @@
>  
>  #define DRV_NAME "ad2s1210"
>  
> +/* The default value of the control register on power-up */
>  #define AD2S1210_DEF_CONTROL		0x7E
>  
> +/* Control Register Bit */
I would change the defines to make this explicit.
This is a truely odd bit of naming anyway.
#define AD2S1210_ADDRESS 0x80
#define AD2S1210_DATA 0x00 
and perhaps a
#define AD2S1210_DATA_MASK 0x7F

would make sense?


>  #define AD2S1210_MSB_IS_HIGH		0x80
>  #define AD2S1210_MSB_IS_LOW		0x7F
>  #define AD2S1210_PHASE_LOCK_RANGE_44	0x20
> @@ -39,14 +41,23 @@
>  
>  #define AD2S1210_REG_POSITION		0x80
>  #define AD2S1210_REG_VELOCITY		0x82
> +
> +/* Loss of Signal (LOS) register address */
>  #define AD2S1210_REG_LOS_THRD		0x88
> +
> +/* Degradation of Signal (DOS) register address */
addresses

>  #define AD2S1210_REG_DOS_OVR_THRD	0x89
>  #define AD2S1210_REG_DOS_MIS_THRD	0x8A
>  #define AD2S1210_REG_DOS_RST_MAX_THRD	0x8B
>  #define AD2S1210_REG_DOS_RST_MIN_THRD	0x8C
> +
> +/* Loss of Tracking (LOT) register address */
addresses 

>  #define AD2S1210_REG_LOT_HIGH_THRD	0x8D
>  #define AD2S1210_REG_LOT_LOW_THRD	0x8E
> +
> +/* Excitation Frequency (EXCIT) register address */
>  #define AD2S1210_REG_EXCIT_FREQ		0x91
> +
>  #define AD2S1210_REG_CONTROL		0x92
>  #define AD2S1210_REG_SOFT_RESET		0xF0
>  #define AD2S1210_REG_FAULT		0xFF
> @@ -69,6 +80,20 @@ enum ad2s1210_mode {
>  
>  static const unsigned int ad2s1210_resolution_value[] = { 10, 12, 14, 16 };
>  
> +/**
> + * struct ad2s1210_state - device instance specific state.
> + * @pdata:		chip model specific constants, gpioin, etc
Except they aren't anything to do with the chip model.  This is about
how it is wired not what it is. 

> + * @lock:		lock to ensure state is consistent
> + * @sdev:		the SPI device for this driver instance
> + * @fclkin:		frequency of clock input
> + * @fexcit:		excitation frequency
> + * @hysteresis:		cache of whether hysteresis is enabled
> + * @old_data:		cache of SPI communication after operation
Umm. You got rid of this one in the earlier patch didn't you?

> + * @resolution:		chip resolution could be 10/12/14/16-bit
>From reading the datasheet quickly I suspect there is a 'best possible'
resolution given a particular set of controls.  I'm not sure we want
to expose this to userspace at all.

> + * @mode:		indicates the operating mode
Where operating mode is what? Comment would be more useful if it
listed them.

> + * @rx:			receive buffer
> + * @tx:			transmit buffer
> + */
>  struct ad2s1210_state {
>  	const struct ad2s1210_platform_data *pdata;
>  	struct mutex lock;
> @@ -82,6 +107,7 @@ struct ad2s1210_state {
>  	u8 tx[2] ____cacheline_aligned;
>  };
>  
> +/* Maps A0 and A1 inputs to the respective mode.  */
>  static const int ad2s1210_mode_vals[4][2] = {
>  	[MOD_POS] = { 0, 0 },
>  	[MOD_VEL] = { 0, 1 },
> @@ -137,6 +163,11 @@ int ad2s1210_update_frequency_control_word(struct ad2s1210_state *st)
>  	int ret;
>  	unsigned char fcw;
>  
> +	/*
> +	 * The fcw stands for frequency control word, which can be obtained
> +	 * from:
> +	 * fcw = (Excitation Frequency * 2^15) / fclkin
> +	 */
Whilst we are here - userspace being responsible for writing a hardware
frequency input needs to change.  Makes no sense.

>  	fcw = (unsigned char)(st->fexcit * (1 << 15) / st->fclkin);
>  	if (fcw < AD2S1210_MIN_FCW || fcw > AD2S1210_MAX_FCW) {
>  		dev_err(&st->sdev->dev, "ad2s1210: FCW out of range\n");
> @@ -158,6 +189,7 @@ static unsigned char ad2s1210_read_resolution_pin(struct ad2s1210_state *st)
>  	return ad2s1210_resolution_value[resolution];
>  }
>  
> +/* Maps RES0 and RES1 inputs to the respective mode.  */
>  static const int ad2s1210_res_pins[4][2] = {
>  	{ 0, 0 }, {0, 1}, {1, 0}, {1, 1}
>  };
> diff --git a/drivers/staging/iio/resolver/ad2s1210.h b/drivers/staging/iio/resolver/ad2s1210.h
> index e9b2147701fc..cbe21bca7638 100644
> --- a/drivers/staging/iio/resolver/ad2s1210.h
> +++ b/drivers/staging/iio/resolver/ad2s1210.h
> @@ -1,5 +1,5 @@
>  /*
> - * ad2s1210.h plaform data for the ADI Resolver to Digital Converters:
> + * ad2s1210.h platform data for the ADI Resolver to Digital Converters:
>   * AD2S1210
Hmm. I would suggest that, seeing as we don't have any in kernel users
we should probably just drop the platform data in favour of a devicetree
binding.

Fair enough to document it as an intermediate step however.
>   *
>   * Copyright (c) 2010-2010 Analog Devices Inc.
> @@ -11,6 +11,13 @@
>  #ifndef _AD2S1210_H
>  #define _AD2S1210_H
>  
> +/**
> + * struct ad2s1210_platform_data - chip model
> + * @sample:	sample input used to clearing the fault register
This hasn't been a good means of proving a gpio for some time.
These all want converting over to the current gpio handling best practice.

> + * @a:		array of inputs (A0 and A1)
> + * @res:	array of resolution inputs (RES0 and RES1)
> + * @gpioin:	control the read operation
In what way?  I think this is actually a hack to allow for the
fact that the above pins may not be controllable by the driver.
Not sure though as I haven't chased through the code fully.

> + */
>  struct ad2s1210_platform_data {
>  	unsigned int		sample;
>  	unsigned int		a[2];

Powered by blists - more mailing lists