On Tue, Jul 19, 2022 at 07:54:18PM +0200, Peter Rosin wrote: > > > 2022-07-19 at 16:44, Chris Morgan wrote: > > On Tue, Jul 19, 2022 at 03:19:01PM +0100, Jonathan Cameron wrote: > >> On Tue, 19 Jul 2022 15:27:24 +0200 > >> Peter Rosin <peda@xxxxxxxxxx> wrote: > >> > >>> Hi! > >>> > >>> 2022-07-19 at 10:51, Jonathan Cameron wrote: > >>>> On Mon, 18 Jul 2022 13:43:12 -0500 > >>>> Chris Morgan <macroalpha82@xxxxxxxxx> wrote: > >>>> > >>>>> From: Chris Morgan <macromorgan@xxxxxxxxxxx> > >>>>> > >>>>> Copy the scan_type details from the parent iio channel to the child. > >>>>> The scan_type is otherwise empty and things like the storagebits are > >>>>> zero (which causes a problem for the adc-joystick driver which > >>>>> validates the storagebits when used through a mux). I'm submitting this > >>>>> as an RFC because I'm not sure if this is the correct way to handle > >>>>> this scenario (a driver that checks the storagebits used with the iio > >>>>> multiplexer). > >>>>> > >>>>> Signed-off-by: Chris Morgan <macromorgan@xxxxxxxxxxx> > >>>> Seems sensible to me. Though Peter is expert on this one. > >>> > >>> I really doubt that it is this simple. I'm no expert on buffered channels, > >>> but after a quick look it certainly looks like adc-joystick requires > >>> buffered channels. And iio-multiplexer does not support that at all. I > >>> think it could be supported, but in that case the only obvious support > >>> would be to support buffered channels one at a time, and the way I > >>> read adc-joystick it expects to get one sample for each axes in one > >>> go (i.e. the way "normal" ADCs deliver multiple parallel samples). See > >>> e.g. the call to iio_channel_get_all_cb in adc-joystick, which dives > >>> right into the buffered iio department. > >>> > >>> So, the stuff in scan_type (along with scan_index, which is highly > >>> related) is only relevant for buffered channels, and iio_multiplexer is > >>> currently limited by its > >>> > >>> indio_dev->modes = INDIO_DIRECT_MODE; > >>> > >>> which is ... not buffered. :-( > >>> > >>> The simplest way forward is probably to lift the requirement of buffered > >>> channels from adc-joystick, but that might be a hard sell as that driver > >>> would then probably be a lot bigger and more complex. > >>> > >> > >> Got it in one ;) > > I guess I'm old and can therefore allow myself to blather :-) > > >> There is a recent series from Chris adding polled support to adc_joystick (not yet merged) > >> https://nam12.safelinks.protection.outlook.com/?url=https%3A%2F%2Flore.kernel.org%2Fall%2F20220705190354.69263-1-macromorgan%40hotmail.com%2F&data=05%7C01%7C%7Ccb22e6634f844f68126e08da69afae8c%7C84df9e7fe9f640afb435aaaaaaaaaaaa%7C1%7C0%7C637938500513929763%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C3000%7C%7C%7C&sdata=00HdVZvnHQNbQ1wFEON%2F9i%2B5b%2BicYb6gRhAJuKooZlA%3D&reserved=0 > >> That just reads the channels one at a time via same interfaces used for > >> sysfs reads so would work with the iio-mux I think. > >> > >> Jonathan > > > > Right, correct that these two "work" together. Honestly the main thing > > is that the adc-joystick driver checks for the storagebits, and that's > > failing for me in this case when I use a mux. I added a one-liner to > > the multiplexer code which added the parent information and that fixed > > it. It's also possible that I could change the adc-joystick driver to > > either not look for the storagebits or to change it to only look when > > not doing polling mode. I just assumed that a missing scan_type was > > an issue, so I added it and marked the patch as RFC because I wasn't > > sure. If you want me to relax the adc-joystick check instead for polled > > devices, I can. > > Ah, ok, there's been new development. Nice! > > As I said, I'm by no means a iio-buffer expert. On the contrary, I'm a > total noob. So, be sure to take the following with a good pinch of salt, > but I think that the adc-joystick driver in a mode that does not use > iio-buffers should not then have artificial requirement on the buffer > layout (scan_type). Don't forget the salt... :-) This sounds like it makes the most sense to me now, since I wasn't aware of the "scan_type" being associated with a buffer. I'll modify the joystick driver and we can disregard this then. Thank you. > > > Note that the adc-joystick driver does work perfectly fine with a mux, > > I'm in the very early stages of upstreaming some devices which use it > > in such a manner (the Anbernic RG353, Anbernic RG503, Odroid Go Super > > are example devices that use the adc-joystick with a mux downstream). > > Cool! > > Cheers, > Peter
