Re: [PATCH 09/10] iio: accel: BMA220 add event attrs

[Jonathan Cameron <jonathan.cameron@huawei.com>]

On Sun, 7 Sep 2025 16:28:17 +0300
Petre Rodan <petre.rodan@subdimension.ro> wrote:

> Hello Jonathan,
> 
> On Sun, Sep 07, 2025 at 02:15:15PM +0100, Jonathan Cameron wrote:
> > On Mon,  1 Sep 2025 22:47:35 +0300
> > Petre Rodan <petre.rodan@subdimension.ro> wrote:
> >   
> > > Add event attributes not directly covered by the IIO API.  
> > 
> > These must be accompanied by ABI documentation in
> > Documentation/ABI/testing/sysfs-bus-iio-...
> > 
> > We need to pull out of the datasheet generic descriptions of what
> > they are so we can consider if they make sense as general new ABI
> > or perhaps map to something existing.  In some cases it is more
> > appropriate just to set a reasonable default and not provide a
> > userspace ABI at all.
> > 
> > Key point is that custom ABI is effectively unused ABI because most
> > software is written against a bunch of devices and has no idea what
> > the new ABI is.   
> 
> I see your point. From all ot those maybe just the interrupt latch functionality
> is generic enough, I have seen something similar while writing other drivers.
> 
> from the datasheet:
> 
> " The interrupt controller can be used in two modes
> 
>  - Latched mode: Once one of the configured interrupt conditions applies, the INT pin is
> asserted and must be reset by the external master through the digital interface.
> 
>  - Non-Latched mode: The interrupt controller clears the INT signal once the interrupt
> condition no longer applies (default behaviour in our chip).
> 
> The interrupt output can be programmed by lat_int[2:0] to be either
>   unlatched ('000') or
>   latched permanently ('111')
>   or have the latch time of 0.25s ('001')/0.5s('010')/1s('011')/2s('100')/4s('101')/8s('110').
> 
> The setting of these bits applies to all types of interrupts."
> 
> Many thanks for the detailed review, I will prepare a new set in b4 and skip
> everything event related for now to keep the set smaller.

Given you are skipping for now we can come back to this later...  But given we
are talking about it I couldn't resist replying ;)

So latched vs unlatched sounds like level signaled interrupt vs edge triggered.

The 'pulse' width cases are annoyingly hard to handle though with those lengths it
seems implausible we'd miss.  I hope for non-latched the sampling rate is slow enough
we still get a good pulse.

Anyhow, I'd suggest ignoring the latch time property and just select between
unlatched and latched based on the interrupt settings in firmware.

Unless you can think of a good usecase for changing these at runtime? I can only
come up with the theory they might be wired to some external circuitry.
Maybe a buzzer to be really annoying ;)

Jonathan

> 
> cheers,
> peter
>