Re: [PATCH v2 1/2] iio: light: add driver for bh1730fvc chips

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



On Thu, 22 Feb 2018 15:58:10 +0100
"Pierre Bourdon (delroth)" <delroth@xxxxxxxxxx> wrote:

> Hi Andy,
> 
> Thanks for the review! Answers inline. I'll send a v3 when the two
> open questions are resolved.
> 
> On Wed, Feb 21, 2018 at 10:57 PM, Andy Shevchenko
> <andy.shevchenko@xxxxxxxxx> wrote:
> > On Wed, Feb 21, 2018 at 9:45 PM, Pierre Bourdon <delroth@xxxxxxxxxx> wrote:  
> >> Ambient light sensor that supports visible light and IR measurements and
> >> configurable gain/integration time.
> >>
> >> Changed in v2:
> >> * Split off DT documentation change into a separate commit.
> >> * Use i2c's probe_new.  
> >
> > Btw, how big the difference with existing drivers?  
> 
> See https://lkml.org/lkml/2018/2/21/982 . In retrospect, I should have
> added this to the commit message. It will be there in v3.
> 
> >> +       highest = max(visible, ir);
> >> +
> >> +       /*
> >> +        * If the read value is being clamped, assume the worst and go to the
> >> +        * lowest possible gain. The alternative is doing multiple
> >> +        * recalibrations, which would be slower and have the same effect.
> >> +        */
> >> +       if (highest == USHRT_MAX)
> >> +               highest *= 128;
> >> +       else
> >> +               highest = (highest * 128) / bh1730_gain_multiplier(bh1730);  
> >
> > In both cases you multiply.
> > Why not just
> >
> >      highest = max(visible, ir) * 128;
> >
> > if (highest < USHRT_MAX)
> > ...
> > ?  
> 
> You'd need < USHRT_MAX * 128 then. I refactored this a bit but
> differently. WDYT?
> 
>         if (highest == USHRT_MAX)
>                 gain = 1;
>         else
>                 gain = bh1730_gain_multiplier(bh1730);
> 
>         highest = (highest * BH1730_MAX_GAIN_MULTIPLIER) / gain;
> 
> >> +       millilux = (u64)USEC_PER_MSEC * (visible_coef * visible - ir_coef * ir);  
> >
> > I'm not sure I understand how time units is related to lux one.  
> 
> Now looks like this, which should match the units together better
> while keeping most multiplications before divisions (for accuracy).
> Not much better I can do here I think, for example the magic "103" is
> straight from the datasheet with no explanation.
> 
>         millilux = visible_coef * visible - ir_coef * ir;
>         millilux = (millilux * USEC_PER_MSEC) / itime_us;
>         millilux *= 103;
>         millilux /= bh1730_gain_multiplier(bh1730);
> 
> >> +       indio_dev = devm_iio_device_alloc(&client->dev, sizeof(*bh1730));
> >> +       if (!indio_dev)
> >> +               return -ENOMEM;
> >> +  
> >  
> >> +       indio_dev->dev.parent = &client->dev;  
> >
> > Strange, it's not done in IIO core... Jonathan, is it true that in
> > case of devm_iio_device_alloc() all drivers use supplied struct device
> > as a parent one?
> > If so, doesn't make sense to modify IIO core to do this?  
> 
> It doesn't seem like IIO core is doing it at this point. Should we
> just leave that as-is for now and get everything fixed in a future
> patch?

It's an interesting point - the bit that bothers me about rolling
that into the devm_iio_device_alloc is that it then
means that function is doing something a little non-obvious given
we don't have the relevant struct device available in the
iio_device_alloc call case.

However, we only have two iio_device_alloc calls left.
The dummy driver one doesn't have a parent (hardly surprising as
there is no hardware).

The other is harder to argue as it's inside the toshiba acpi
driver where reworking to make devm possible would be a major
job.

Hmm.  Will think about it...

> 
> >> +static int bh1730_remove(struct i2c_client *client)
> >> +{
> >> +       struct iio_dev *indio_dev = i2c_get_clientdata(client);
> >> +       struct bh1730_data *bh1730 = iio_priv(indio_dev);
> >> +  
> >  
> >> +       iio_device_unregister(indio_dev);  
> >
> > Hmm... Do you still need this even with devm IIO in ->probe()?  
> 
> I *think* so, but that's mostly because all the other IIO light
> drivers that I've looked at are doing it. Can someone who knows the
> IIO subsystem well weigh in?

Yes, it is absolutely needed.  This removes the userspace and core
kernel interfaces 'before' we turn the power off.  Without it there
are some obvious race conditions.  It's hard to query a sensor
which is powered down after all.

You could wrap up the power off call in devm magic, but that will
probably be more complex and error prone than doing it as we have
here.

In cases where there is no clean up to do in order we can
move to fully managed - this isn't one of those though.

Jonathan

> 
> Thanks!
> Best,
> 

--
To unsubscribe from this list: send the line "unsubscribe devicetree" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html



[Index of Archives]     [Device Tree Compilter]     [Device Tree Spec]     [Linux Driver Backports]     [Video for Linux]     [Linux USB Devel]     [Linux PCI Devel]     [Linux Audio Users]     [Linux Kernel]     [Linux SCSI]     [XFree86]     [Yosemite Backpacking]


  Powered by Linux