Re: [PATCH 3/3] mfd: arizona: Use regmap_read_poll_timeout instead of hard coding it

[Lee Jones]

On Thu, 05 Jan 2017, Charles Keepax wrote:

> On Thu, Jan 05, 2017 at 08:07:01AM +0000, Lee Jones wrote:
> > On Wed, 04 Jan 2017, Charles Keepax wrote:
> > 
> > > arizona_poll_reg essentially hard-codes regmap_read_poll_timeout, this
> > > patch updates the implementation to use regmap_read_poll_timeout. We
> > > still keep arizona_poll_reg around as regmap_read_poll_timeout is a
> > > macro so rather than expand this for each caller keep it wrapped in
> > > arizona_poll_reg.
> > > 
> > > Signed-off-by: Charles Keepax <ckeepax@xxxxxxxxxxxxxxxxxxxxxxxxxxx>
> > > ---
> > >  drivers/mfd/arizona-core.c | 28 +++++++++++-----------------
> > >  1 file changed, 11 insertions(+), 17 deletions(-)
> > > 
> > > diff --git a/drivers/mfd/arizona-core.c b/drivers/mfd/arizona-core.c
> > > index 4cb34c3..e6fae3c 100644
> > > --- a/drivers/mfd/arizona-core.c
> > > +++ b/drivers/mfd/arizona-core.c
> > > @@ -236,28 +236,22 @@ static irqreturn_t arizona_overclocked(int irq, void *data)
> > >  }
> > >  
> > >  static int arizona_poll_reg(struct arizona *arizona,
> > > -			    int timeout, unsigned int reg,
> > > +			    int npolls, unsigned int reg,
> > >  			    unsigned int mask, unsigned int target)
> > >  {
> > > +	const int poll_us = 7500;
> > 
> > Get rid of this and replace its usage with a nice #define describing
> > exactly what the timeout is for i.e what timed out.
> > 
> 
> I can replace this with a define if you prefer although since the
> value is only used locally I generally prefer a const variable as
> it keeps all the relevant code together.

I'd prefer a DEFINE, then you can rid the requirement for a variable
altogether.

> > ... okay, I just read the doc.  This should really be 'SLEEP'.
> > 
> > Where did 7500 come from anyway?  Is it documented?
> > 
> 
> To be fair this number is a little pulled from my ... err the
> air. But basically provides a reasonable balance of timing
> between polls for the various users of the function. It didn't
> really seem worth having each user specify its own timing for
> this part as we generally don't care down to a couple of mS here
> or there.

Very well.  If you define it I'll ask no more questions. ;)

> > >  	unsigned int val = 0;
> > > -	int ret, i;
> > > -
> > > -	for (i = 0; i < timeout; i++) {
> > > -		ret = regmap_read(arizona->regmap, reg, &val);
> > > -		if (ret != 0) {
> > > -			dev_err(arizona->dev, "Failed to read reg 0x%x: %d\n",
> > > -				reg, ret);
> > > -			continue;
> > > -		}
> > > -
> > > -		if ((val & mask) == target)
> > > -			return 0;
> > > +	int ret;
> > >  
> > > -		usleep_range(1000, 5000);
> > > -	}
> > > +	ret = regmap_read_poll_timeout(arizona->regmap,
> > > +				       ARIZONA_INTERRUPT_RAW_STATUS_5, val,
> > > +				       ((val & mask) == target), poll_us,
> > > +				       npolls * poll_us);
> > 
> > What's the relevance of npolls?  Is this number documented, or was it
> > pulled from your ... err, the air?
> > 
> 
> Its an argument passed into the function, its called npolls in a
> function that polls a register I really don't know how to make
> the fact it is the number of polls to do any more clear than

I can see what it's used for, I am more concerned about the number
passed in.

> that. As for the number passed into the function that depends on
> how long that particular register should be polled for, this is a
> function that is used in several places for polling for a
> particular register state.

I would have thought a consumer would be more likely to know how long
it would poll for, rather than how many times to poll.  Knowledge of
time-between-polls in only known locally.  Do you see where I'm going
with this?

> Would all these issues perhaps be best solved by my adding some
> kernel doc for the function with some more description of what
> everything is?

I don't think so, to be honest.

> > > +	if (ret)
> > > +		dev_err(arizona->dev, "Polling reg 0x%x timed out: %x\n",
> > > +			reg, val);
> > >  
> > > -	dev_err(arizona->dev, "Polling reg 0x%x timed out: %x\n", reg, val);
> > > -	return -ETIMEDOUT;
> > > +	return ret;
> > >  }
> > >  
> > >  static int arizona_wait_for_boot(struct arizona *arizona)
> > 

-- 
Lee Jones
Linaro STMicroelectronics Landing Team Lead
Linaro.org â Open source software for ARM SoCs
Follow Linaro: Facebook | Twitter | Blog