Re: [PATCH v8 2/2] iio: adc: Add ti-ads1018 driver

["Kurt Borja" <kuurtb@gmail.com>]

On Tue Dec 16, 2025 at 1:21 PM -05, Jonathan Cameron wrote:
> On Mon, 15 Dec 2025 12:09:31 -0600
> David Lechner <dlechner@baylibre.com> wrote:
>
>> On 12/15/25 10:54 AM, Kurt Borja wrote:
>> > On Mon Dec 15, 2025 at 10:55 AM -05, David Lechner wrote:  
>> >> On 12/11/25 10:25 PM, Kurt Borja wrote:  
>> >>> Add ti-ads1018 driver for Texas Instruments ADS1018 and ADS1118 SPI
>> >>> analog-to-digital converters.
>> >>>  
>> 
>> ...
>> 
>> >>> +static irqreturn_t ads1018_trigger_handler(int irq, void *p)
>> >>> +{
>> >>> +	struct iio_poll_func *pf = p;
>> >>> +	struct iio_dev *indio_dev = pf->indio_dev;
>> >>> +	struct ads1018 *ads1018 = iio_priv(indio_dev);
>> >>> +	struct {
>> >>> +		__be16 conv;
>> >>> +		aligned_s64 ts;
>> >>> +	} scan = {};
>> >>> +	int ret;
>> >>> +  
>> >>
>> >>  
>> >>> +	if (iio_device_claim_buffer_mode(indio_dev))
>> >>> +		goto out_notify_done;  
>> >>
>> >> This should not be needed. It should not be possible to
>> >> exit buffer mode without triggers being stopped first.
>> >> (No other driver is doing this.)  
>> > 
>> > Previously I had my own lock here because ads1018_spi_read_exclusive()
>> > needs locking.  
>> 
>> What exactly are we protecting against here? I.e. give side-by-side
>> lists of possible concurrent function calls where there could be a
>> problem.
>> 
>> Any call to iio_device_claim_direct() will already fail without
>> calling iio_device_claim_buffer_mode() here. And since this is
>> an interrupt handler, we don't have to worry about reentrancy (it
>> can't be called again until the previous call returns). And nowhere
>> else in the driver is calling iio_device_claim_buffer_mode(). So
>> calling it here doesn't actually add any protection AFAICT.

You're right. I assumed threaded IRQs were reentrant, which doesn't seem
to be the case. Thanks for pointing it out!

>
> Agreed we shouldn't need this. Given these comment and my lazy nature,
> Kurt, would you mind spinning a patch on top of this series that I can
> squash with it on my tree?  That should be easier to review than
> a full v9.  If you prefer a v9 of the whole thing, that would be fine too.

Sure! I'll send a patch later today.

>
> Thanks,
>
> Jonathan

-- 
Thanks,
 ~ Kurt