On Tue, 30 Mar 2021 21:23:59 +0000 "Song Bao Hua (Barry Song)" <song.bao.hua@xxxxxxxxxxxxx> wrote: > > -----Original Message----- > > From: Jonathan Cameron [mailto:jic23@xxxxxxxxxxxxxxxxxxxxx] > > Sent: Tuesday, March 30, 2021 4:36 AM > > To: linux-iio@xxxxxxxxxxxxxxx; Song Bao Hua (Barry Song) > > <song.bao.hua@xxxxxxxxxxxxx> > > Cc: Lars-Peter Clausen <lars@xxxxxxxxxx>; Michael Hennerich > > <Michael.Hennerich@xxxxxxxxxx>; Alexandru Ardelean <aardelean@xxxxxxxxxxx>; > > Robh+dt@xxxxxxxxxx; Alexandru Ardelean <ardeleanalex@xxxxxxxxx>; Jonathan > > Cameron <jonathan.cameron@xxxxxxxxxx> > > Subject: Re: [PATCH v2 00/24] staging:iio:cdc:ad7150: cleanup / fixup / graduate > > > > On Mon, 29 Mar 2021 16:30:21 +0100 > > Jonathan Cameron <jic23@xxxxxxxxxx> wrote: > > > > > Hi All, > > > > > > Whilst I'll give up at some point and just apply this without additional > > > tags I really don't like doing that as I've made too many idiot mistakes > > > in the past. > > > > > > Some of these are fairly trivial so please can people take a look if > > > they have a chance. Rob did the DT one (thanks!) so we are down to... > > > > > > > > > > > 12 - The big irq rework patch. Alex wasn't quite happy to give a tag > > > > on that last time, but didn't mention anything specific. It's a bit > > > > fiddly so fair enough! > > > > > > (it's not that bad!) > > > > > > > 20 had an ack that I think was accidentally sent off list and hence burried > > in my junk. And then there were two... > > > > > > > > > > 21 - ABI patch. I don't think there is anything controversial in here > > > > but it's not gained any tags yet. > > > > > > > > > > Perhaps didn't help that I accidentally didn't cc Barry on v2. > > > > > > Thanks, > > > > > > Jonathan > > > > > > > > > > > > > v1 description: > > > > > > > > This is an 'old' driver in IIO that has been in staging a while. > > > > First submitted in October 2010. > > > > > > > > I wanted to try and experiment and picked this driver to try it with. > > > > > > > > The cleanup etc here was all tested against some basic emulation > > > > added to QEMU rather than real hardware. Once I've cleaned that up > > > > a tiny bit I'll push it up to https://github.com/jic23/qemu > > > > Note that for now I'm not proposing to upstream this to QEMU but > > > > would be interested in hearing if people thing it is a good idea to > > > > do so. > > > > > > > > Whilst it's obviously hard to be absolutely sure that the emulation is > > > > correct, the fact that the original driver worked as expected and the > > > > cleaned up version still does is certainly encouraging. > > > > > > > > Note however, that there were a few more significant changes in here than > > > > basic cleanup. > > > > 1. Interrupts / events were handled in a rather esoteric fashion. > > > > (Always on, window modes represented as magnitudes). > > > > Note that for two channel devices there are separate lines. The original > > > > driver was not supporting this at all. > > > > They now look more like a standard IIO driver and reflect experience > > > > that we've gained over the years in dealing with devices where these > > > > aren't interrupt lines as such, but rather reporters of current status. > > > > 2. Timeouts were handled in a fashion that clearly wouldn't work. > > > > > > > > Note that this moving out of staging makes a few bits of ABI 'official' > > > > and so those are added to the main IIO ABI Docs. > > > > > > > > Thanks in advance to anyone who has time to take a look. > > > > > > > > Jonathan Cameron (24): > > > > staging:iio:cdc:ad7150: use swapped reads for i2c rather than open > > > > coding. > > Using i2c_smbus_read_word_swapper and i2c_smbus_write_word_swapped > looks good to me. The only thing is that your changelog didn't > mention you also used swapper write as you said "use swapped > reads". Otherwise, > > Reviewed-by: Barry Song <song.bao.hua@hisilicon> > > > > > staging:iio:cdc:ad7150: Remove magnitude adaptive events > > > > staging:iio:cdc:ad7150: Refactor event parameter update > > > > staging:iio:cdc:ad7150: Timeout register covers both directions so > > > > both need updating > > > > staging:iio:cdc:ad7150: Drop platform data support > > > > staging:iio:cdc:ad7150: Handle variation in chan_spec across device > > > > and irq present or not > > > > staging:iio:cdc:ad7150: Simplify event handling by only using rising > > > > direction. > > > > staging:iio:cdc:ad7150: Drop noisy print in probe > > > > staging:iio:cdc:ad7150: Add sampling_frequency support > > > > iio:event: Add timeout event info type > > > > staging:iio:cdc:ad7150: Change timeout units to seconds and use core > > > > support > > > > staging:iio:cdc:ad7150: Rework interrupt handling. > > + /* > + * There are race conditions around enabling and disabling that > + * could easily land us here with a spurious interrupt. > + * Just eat it if so. > + */ > + if (!(int_status & status_mask)) > + return IRQ_HANDLED; > + > > I am not sure what kind of race conditions we have since disable_irq() > will synchronize with the irq handler. > > If we are in an interrupt, the user who calls disable_irq will wait > for the completion of irq handler. > If an interrupt comes in the gap of disable_irq and enable_irq, we should > have a valid int_status after we re-enable the interrupt? > > Maybe it is because of the weird behavior of the hardware? I think you are right and I was being overly paranoid on that one. I'll drop the comment, though I'll leave the paranoid check. > > > > > staging:iio:cdc:ad7150: More consistent register and field naming > > > > staging:iio:cdc:ad7150: Reorganize headers. > > > > staging:iio:cdc:ad7150: Tidy up local variable positioning. > > > > staging:iio:cdc:ad7150: Drop unnecessary block comments. > > > > staging:iio:cdc:ad7150: Shift the _raw readings by 4 bits. > > > > staging:iio:cdc:ad7150: Add scale and offset to > > > > info_mask_shared_by_type > > > > staging:iio:cdc:ad7150: Really basic regulator support. > > > > staging:iio:cdc:ad7150: Add of_match_table > > Reviewed-by: Barry Song <song.bao.hua@hisilicon> > > > > > iio:Documentation:ABI Add missing elements as used by the adi,ad7150 > > +What: /sys/.../events/in_capacitanceY_adaptive_thresh_rising_en > +What: /sys/.../events/in_capacitanceY_adaptive_thresh_falling_en > +KernelVersion: 5.11 > > Is kernel 5.11 the right version? Guess not :-) Oops Can tell this took a while. I'll tidy that up whilst applying. > > > > > staging:iio:cdc:ad7150: Add copyright notice given substantial > > > > changes. > > > > dt-bindings:iio:cdc:adi,ad7150 binding doc > > Reviewed-by: Barry Song <song.bao.hua@hisilicon> > > + properties: > + compatible: > + contains: > + enum: > + - adi,ad7150 > + - adi,ad7156 > + then: > + properties: > + interrupts: > + minItems: 2 > + maxItems: 2 > + - if: > + properties: > + compatible: > + contains: > + const: adi,ad7151 > + then: > + properties: > + interrupts: > + minItems: 1 > + maxItems: 1 > > A further follow-up might be: > we can move to read two irq number for adi7150/7156 > and one irq number for adi7151 in driver? Unless I'm miss understanding you I think we already do. There is a check against the driver_data being AD7150 which is set for the two parts that have 2 interrupts. If we don't get that we don't query the second irq number. > > > > > iio:cdc:ad7150: Move driver out of staging. > > > > > > > > Documentation/ABI/testing/sysfs-bus-iio | 33 + > > > > .../bindings/iio/cdc/adi,ad7150.yaml | 69 ++ > > > > drivers/iio/Kconfig | 1 + > > > > drivers/iio/Makefile | 1 + > > > > drivers/iio/cdc/Kconfig | 17 + > > > > drivers/iio/cdc/Makefile | 6 + > > > > drivers/iio/cdc/ad7150.c | 678 ++++++++++++++++++ > > > > drivers/iio/industrialio-event.c | 1 + > > > > drivers/staging/iio/cdc/Kconfig | 10 - > > > > drivers/staging/iio/cdc/Makefile | 3 +- > > > > drivers/staging/iio/cdc/ad7150.c | 655 ----------------- > > > > include/linux/iio/types.h | 1 + > > > > 12 files changed, 808 insertions(+), 667 deletions(-) > > > > create mode 100644 > > Documentation/devicetree/bindings/iio/cdc/adi,ad7150.yaml > > > > create mode 100644 drivers/iio/cdc/Kconfig > > > > create mode 100644 drivers/iio/cdc/Makefile > > > > create mode 100644 drivers/iio/cdc/ad7150.c > > > > delete mode 100644 drivers/staging/iio/cdc/ad7150.c > > > > > > > > > > Thanks > Barry > Thanks! Jonathan
