[PATCH v5 0/2] iio: adc: Add support for Texas Instruments ADS112C04
Kyle Hsieh <[email protected]>
| Newsgroups | org.kernel.vger.linux-devicetree,org.kernel.vger.linux-iio,org.kernel.vger.linux-kernel |
|---|---|
| Message-ID | <[email protected]> |
This patch series introduces support for the Texas Instruments ADS112C04 Analog-to-Digital Converters. The ADS112C04 (16-bit) is precision, low-power, delta-sigma ADCs with an I2C interface. They feature a flexible input multiplexer supporting single-ended and differential measurements, a programmable gain amplifier, and an internal voltage reference. Note: While this chip shares similarities with the ADS112C14 (currently being upstreamed by David Lechner), the register maps and feature sets are sufficiently different to warrant a separate driver. However, the DT bindings and channel parsing logic have been aligned with the ADS112C14 conventions. This initial submission provides a minimal feature set (single-shot conversions and basic DRDY interrupt) covering current use cases. Signed-off-by: Kyle Hsieh <[email protected]> --- Changes in v5: - Addressed several suggestions from v1-v4 that I had missed in earlier revisions. Apologies for the churn this caused; I went back through all four threads as David asked. - dt-bindings: iio: adc: ti,ads112c04: - No changes. Picked up David's Reviewed-by and Conor's Acked-by. - iio: adc: ti-ads112c04: - Made the reference voltage per-channel, following the ADS112C14 driver: reference-sources is parsed with fwnode_property_match_property_string() and stored per channel, and the CONFIG1 VREF field is now updated together with the MUX before each conversion. avdd is read with devm_regulator_get_enable_read_voltage() when a channel uses it. - Organised the register field macros indented under their register, and shortened CONFIG to CONF in the field names. - Spelled out all CONFIG0/CONFIG1 fields explicitly with FIELD_PREP() when initialising, rather than relying on implicit zeros. - Moved the internal reference macro to the top of the file with the other defines and renamed it to ADS112C04_INT_REF_mV. - Dropped disable_irq_nosync(); the handler only calls complete(), which is safe in a normal IRQ handler, so a threaded IRQ isn't needed. The bare `0` flags argument is now packed with the other arguments. - Fixed word-read byte ordering by using i2c_smbus_read_word_swapped(). - Dropped the has_refp state member; it is passed as a function argument instead. - Added a REVISIT comment on spec->type for when ti,refp-refn-resistor-ohms is implemented. - Reduced the power-on reset delay from 50ms to 500us, per the datasheet. - Replaced the diff-channels if-else chain with a 4x4 lookup table indexed by [AINP][AINN]. - Used wait_for_completion_timeout() directly in the if condition, swapped err/ret in ads112c04_wait_for_data(), made the channel index unsigned int with the assignment split out, stopped reusing pair[0] for single-channel, and used ARRAY_SIZE(pair). - Added ADS112C04_MAX_CHANNELS and used it in the bounds check. - Split reference-sources parsing into a helper to keep the call within the line length limit. - Updated the commit message to describe per-channel reference selection; refn-supply remains unsupported. - Switched the hardware reset to the reset controller framework (devm_reset_control_get_optional_exclusive()), which works with the existing reset-gpios property via the reset-gpio driver. The binding is unchanged. - Link to v4: https://lore.kernel.org/r/[email protected] Changes in v4: - dt-bindings: iio: adc: ti,ads112c04: - Fixed excitation-current-nanoamp to use single-entry list syntax (items: - enum: [...]) instead of combining maxItems: 1 with a mapping-style items, which failed dt_binding_check. - Widened excitation-channels to maximum: 5 to account for IDAC1/ IDAC2 also being routable to REFP0/REFN0, and documented the mapping, per David Lechner. - Reduced excitation-current-nanoamp to a single entry, since the two IDAC outputs share one current setting, per David Lechner. - Changed burn-out-current-nanoamp to use `const: 10000`, per David Lechner. - Renamed reference-sources' internal enum value to "internal" (dropped voltage suffix, since there is only one internal reference), per David Lechner. - iio: adc: ti-ads112c04: - Fixed word-read byte ordering by switching to i2c_smbus_read_word_swapped(), addressing a big-endian data corruption issue flagged by Sashiko AI review. - Masked the DRDY interrupt in the hardirq handler (disable_irq_nosync()) and re-enabled it once the conversion data is read, to avoid an interrupt storm on level-triggered DRDY configurations, without hardcoding the trigger type. - Reworked reference-sources handling in parse_channels(): it is now validated against the actually-selected refp-supply instead of being unconditionally rejected, per David Lechner. - Simplified regulator handling in probe() using devm_regulator_get_enable_read_voltage(), dropping the now-unused vref_reg field, per David Lechner. - Corrected the commit message to no longer claim refn-supply support, since it remains explicitly unsupported in probe(). - Added a MAINTAINERS entry for the new binding and driver. - Link to v3: https://lore.kernel.org/r/[email protected] Changes in v3: - dt-bindings: iio: adc: ti,ads112c04: - Resolved dt_binding_check errors by removing redundant $ref for standard unit suffixes and fixing YAML array syntax. - Added full hardware capability descriptions (excitation-channels, excitation-current-nanoamp, burn-out-current-nanoamp, reference-sources) as suggested by David Lechner. - Restricted the reg maximum to 11 and updated the regex to ^channel@[0-9a-b]$ to accurately reflect the 12 possible MUX combinations. - iio: adc: ti-ads112c04: - Transitioned all I2C read/write wrappers to use SMBus APIs (i2c_smbus_read_byte_data, i2c_smbus_read_word_data, etc.) to gracefully handle I2C errors, NACKs, and Repeated Starts, addressing feedback from Joshua Crofts and Jonathan Cameron. - Added forward compatibility checks in probe() and parse_channels() using dev_err_probe() to return -EOPNOTSUPP for unimplemented DT properties. - Reverted devm_request_irq() flags to 0 to let the driver inherit the trigger type strictly from the DT, per Jonathan Cameron's advice. - Reduced the data wait timeout to 100ms, reflecting the chip's slowest data rate of 20 SPS. - Fixed #include alphabetical sorting, grouped <linux/iio/*> headers, and added missing headers. - Fixed variable declaration ordering (reverse xmas tree) and updated variable naming to vref_mV (SI unit exception). - Optimized GPIO reset logic using GPIOD_OUT_HIGH. - Link to v2: https://lore.kernel.org/r/[email protected] Changes in v2: - Replaced `vref-supply` with `refp-supply` and `refn-supply` to accurately reflect hardware. - Refactored the driver to dynamically parse channel configurations and routing from DT child nodes. - Modernized the driver using kernel macros. - Handled endianness elegantly. - Added hardware reset fallback logic. - Inherited IRQ trigger type from device tree instead of hardcoding. - Fixed a bug where the MUX software cache could desync from hardware if the I2C write failed. - Added strict return value checking for all I2C writes during probe. - Updated the `i2c_device_id` array to use C99 named initializers. - Link to v1: https://lore.kernel.org/r/[email protected] --- Kyle Hsieh (2): dt-bindings: iio: adc: ti,ads112c04: Add binding for ADS112C04 iio: adc: ti-ads112c04: Add support for TI ADS112C04 .../devicetree/bindings/iio/adc/ti,ads112c04.yaml | 148 ++++++ MAINTAINERS | 7 + drivers/iio/adc/Kconfig | 10 + drivers/iio/adc/Makefile | 1 + drivers/iio/adc/ti-ads112c04.c | 525 +++++++++++++++++++++ 5 files changed, 691 insertions(+) --- base-commit: 350d1fb9204b13c5f95e511e98b8bcb47574d425 change-id: 20260724-ti-ads112c04-driver-be7e89047834 Best regards, -- Kyle Hsieh <[email protected]>