RE: [PATCH v6 0/8] iio: imu: Add support for the ADI ADIS16607
From: Sabau, Radu bogdan
Date: Mon Nov 16 2026 - 07:18:54 EST
> -----Original Message-----
> From: Sabau, Radu bogdan
> Sent: Monday, November 16, 2026 10:55 AM
...
> > >
> > > > This series adds support for the Analog Devices ADIS16607, a precision
> > > > MEMS IMU with triaxial gyroscope, triaxial accelerometer, and
> > > > temperature sensor, communicating over a non-standard SPI Half-Duplex
> > > > protocol.
> > >
> > > Applied 1-5 to the testing branch of iio.git.
> > >
> > > I 'think' the proposed changes for the fifo handling shouldn't affect
> > > those and I don't fancy reviewing a mega series with library changes
> > > and the new device support. So build on the base of these first 5 patches.
> > >
> > > You will need to provide a base commit for dependency list though
> > > or sashiko may get confused.
> > >
> > Change of plan. Dropped it over channel type issues in patch 5.
>
> Thank you guys again for the review, and sorry about patch 5. I'll move the
> scan_index/scan_type and the timestamp channel into the FIFO patch and
> resend.
>
> On the FIFO side: adis16475's has_fifo variants register an IIO trigger
> today and are used through trigger/current_trigger. Converting them to a
> trigger-less kfifo buffer, as discussed for the library, would remove
> that attribute on the six ADIS1657x variants, so it is an ABI change on
> existing devices. Is that acceptable to you?
>
> If not, the alternative is a smaller fix in adislib: iio_trigger_poll_nested()
> for the has_fifo path, plus a prep patch taking the sample timestamp in
> the handlers instead of pf->timestamp (the nested path does not run the
> pollfunc top half). That keeps the ABI as is, and the adis16607 would
> then use the adislib trigger path again.
Thinking about how to split this: sync and calibration bias do not depend
on the FIFO, so I could resend the series as the adislib patches, the
binding, the base driver (raw reads only), external clock sync and
calibration bias, with no FIFO. Then a follow-up series with the FIFO.
Would that work for you?
For that follow-up I would still like to know which way you prefer to
go for the adislib FIFO handling:
- kfifo helper: a trigger-less kfifo buffer with its own threaded IRQ,
used by the adis16607 and the adis16475 FIFO variants. It removes the
trigger/current_trigger attribute on those variants, so it is an ABI
change.
- iio_trigger_poll_nested(): a smaller fix in adislib for the has_fifo
path, plus a prep patch taking the sample timestamp in the handlers.
It keeps the ABI, and the adis16607 would use the trigger path again.
It is only two or three patches, so if you think it is the better
alternative it could also be done in this series.